Reconciled evidence

The agency reconciles every validated claim against the others. Each relationship below is persisted, not inferred in the browser: it records how the findings relate, which context dimensions differ or are unknown, the agency’s interpretation, and the confidence adjustment that follows.

2402 relationships · 100 claims resolved

Contextually different (not comparable)

2155

Measured differently, so the numbers should not be read as the same quantity.

Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The Grok 4.7 model supports a context window of 500,000 tokens.

context window
500000.0 tokens

Compared metric: m1 — context window

xAIsourceoptimisation implicationevidence for grok

The Responses API for agentic coding with Grok has an early access output token cost of $6.00 per 1 million tokens.

early access cost per 1 million output t
6.0 · 1M output tokens

Compared metric: m1 — early access cost per 1 million output t

xAIsourceoptimisation implicationevidence for grok

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The Grok 4.7 model supports a context window of 500,000 tokens.

context window
500000.0 tokens

Compared metric: m1 — context window

xAIsourceoptimisation implicationevidence for grok

The Responses API for agentic coding with Grok has an early access input token cost of $2.00 per 1 million tokens.

early access cost per 1 million input to
2.0 · 1M input tokens · Early access

Compared metric: m1 — early access cost per 1 million input to

xAIsourceoptimisation implicationevidence for grok

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The Grok 4.7 model supports a context window of 500,000 tokens.

context window
500000.0 tokens

Compared metric: m1 — context window

xAIsourceoptimisation implicationevidence for grok

The Grok 4.7 model has an output token cost of $6.00 per 1 million tokens.

cost per 1 million output tokens
6.0 · 1M tokens · Output

Compared metric: m1 — cost per 1 million output tokens

xAIsourceoptimisation implicationevidence for grok

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The Grok 4.7 model supports a context window of 500,000 tokens.

context window
500000.0 tokens

Compared metric: m1 — context window

xAIsourceoptimisation implicationevidence for grok

The Grok 4.7 model has an input token cost of $2.00 per 1 million tokens.

cost per 1 million input tokens
2.0 · 1M tokens · Input

Compared metric: m1 — cost per 1 million input tokens

xAIsourceoptimisation implicationevidence for grok

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The Responses API for agentic coding with Grok has an early access output token cost of $6.00 per 1 million tokens.

early access cost per 1 million output t
6.0 · 1M output tokens

Compared metric: m1 — early access cost per 1 million output t

xAIsourceoptimisation implicationevidence for grok

The Responses API for agentic coding with Grok has an early access input token cost of $2.00 per 1 million tokens.

early access cost per 1 million input to
2.0 · 1M input tokens · Early access

Compared metric: m1 — early access cost per 1 million input to

xAIsourceoptimisation implicationevidence for grok

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The Responses API for agentic coding with Grok has an early access output token cost of $6.00 per 1 million tokens.

early access cost per 1 million output t
6.0 · 1M output tokens

Compared metric: m1 — early access cost per 1 million output t

xAIsourceoptimisation implicationevidence for grok

The Grok 4.7 model has an input token cost of $2.00 per 1 million tokens.

cost per 1 million input tokens
2.0 · 1M tokens · Input

Compared metric: m1 — cost per 1 million input tokens

xAIsourceoptimisation implicationevidence for grok

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The Responses API for agentic coding with Grok has an early access input token cost of $2.00 per 1 million tokens.

early access cost per 1 million input to
2.0 · 1M input tokens · Early access

Compared metric: m1 — early access cost per 1 million input to

xAIsourceoptimisation implicationevidence for grok

The Grok 4.7 model has an output token cost of $6.00 per 1 million tokens.

cost per 1 million output tokens
6.0 · 1M tokens · Output

Compared metric: m1 — cost per 1 million output tokens

xAIsourceoptimisation implicationevidence for grok

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The Grok 4.7 model has an output token cost of $6.00 per 1 million tokens.

cost per 1 million output tokens
6.0 · 1M tokens · Output

Compared metric: m1 — cost per 1 million output tokens

xAIsourceoptimisation implicationevidence for grok

The Grok 4.7 model has an input token cost of $2.00 per 1 million tokens.

cost per 1 million input tokens
2.0 · 1M tokens · Input

Compared metric: m1 — cost per 1 million input tokens

xAIsourceoptimisation implicationevidence for grok

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google employs thousands of quality raters to test and improve search and AI results.

number of quality raters
1000.0 thousands · Google's quality raters

Compared metric: m1 — number of quality raters

Search Engine Roundtablesourceretrieval indexevidence for google-ai-mode

AI Overviews in France cut the click-through rate (CTR) by 23.1%.

CTR reduction
23.1 percent · in France

Compared metric: m1 — CTR reduction

Ahrefssourcereferrals conversionevidence for google-ai-mode

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google employs thousands of quality raters to test and improve search and AI results.

number of quality raters
1000.0 thousands · Google's quality raters

Compared metric: m1 — number of quality raters

Search Engine Roundtablesourceretrieval indexevidence for google-ai-mode

Google's AI Mode has over one billion monthly users, one year after its launch.

monthly users
1000000000.0 · one year after launch · AI Mode

Compared metric: m1 — monthly users

SISTRIXsourceaudience usageevidence for google-ai-mode

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI Overviews in France cut the click-through rate (CTR) by 23.1%.

CTR reduction
23.1 percent · in France

Compared metric: m1 — CTR reduction

Ahrefssourcereferrals conversionevidence for google-ai-mode

Google's AI Mode has over one billion monthly users, one year after its launch.

monthly users
1000000000.0 · one year after launch · AI Mode

Compared metric: m1 — monthly users

SISTRIXsourceaudience usageevidence for google-ai-mode

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Screaming Frog SEO Crawler averages about 2 URLs per second for crawling and about 0.5 URLs per second when analyzed by Gemini 3.5 Flash Lite.

crawling rate
2.0
analysis rate by Gemini 3.5 Flash Lite
0.5

Compared metric: m1 — crawling rate

SISTRIXsourcemeasurementevidence for gemini

Screaming Frog SEO Crawler averages about 2 URLs per second for crawling and about 0.5 URLs per second when analyzed by Gemini 3.5 Flash Lite.

crawling rate
2.0
analysis rate by Gemini 3.5 Flash Lite
0.5

Compared metric: m2 — analysis rate by Gemini 3.5 Flash Lite

SISTRIXsourcemeasurementevidence for gemini

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Screaming Frog SEO Crawler averages about 2 URLs per second for crawling and about 0.5 URLs per second when analyzed by Gemini 3.5 Flash Lite.

crawling rate
2.0
analysis rate by Gemini 3.5 Flash Lite
0.5

Compared metric: m1 — crawling rate

SISTRIXsourcemeasurementevidence for gemini

The Gemini 3.5 Flash Lite model can run faster than the crawling process, processing approximately 0.5 URLs per second.

URLs analyzed per second
0.5 URLs/second · Gemini 3.5 Flash Lite

Compared metric: m1 — URLs analyzed per second

SISTRIXsourceretrieval indexevidence for gemini

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Screaming Frog SEO Crawler averages about 2 URLs per second for crawling and about 0.5 URLs per second when analyzed by Gemini 3.5 Flash Lite.

crawling rate
2.0
analysis rate by Gemini 3.5 Flash Lite
0.5

Compared metric: m2 — analysis rate by Gemini 3.5 Flash Lite

SISTRIXsourcemeasurementevidence for gemini

The Gemini 3.5 Flash Lite model can run faster than the crawling process, processing approximately 0.5 URLs per second.

URLs analyzed per second
0.5 URLs/second · Gemini 3.5 Flash Lite

Compared metric: m1 — URLs analyzed per second

SISTRIXsourceretrieval indexevidence for gemini

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-rate of a search with an AI Overview is reduced from 27% to 11% at position 1.

click-rate without AI Overview
27.0
click-rate with AI Overview
11.0

Compared metric: m1 — click-rate without AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

The click-rate of a search with an AI Overview is reduced from 27% to 11% at position 1.

click-rate without AI Overview
27.0
click-rate with AI Overview
11.0

Compared metric: m2 — click-rate with AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-rate of a search with an AI Overview is reduced from 27% to 11% at position 1.

click-rate without AI Overview
27.0
click-rate with AI Overview
11.0

Compared metric: m1 — click-rate without AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

Google's AI Overviews have significantly increased their inclusion of external links, rising from nearly 0% to over 25% of AI Overviews.

percentage of AI Overviews containing external links
0.0 % · AI Overviews
percentage of AI Overviews containing external links
25.0 % · AI Overviews

Compared metric: m1 — percentage of AI Overviews containing external links

Search Engine Roundtablesourcecitations sourcesevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-rate of a search with an AI Overview is reduced from 27% to 11% at position 1.

click-rate without AI Overview
27.0
click-rate with AI Overview
11.0

Compared metric: m1 — click-rate without AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

Google's AI Overviews have significantly increased their inclusion of external links, rising from nearly 0% to over 25% of AI Overviews.

percentage of AI Overviews containing external links
0.0 % · AI Overviews
percentage of AI Overviews containing external links
25.0 % · AI Overviews

Compared metric: m2 — percentage of AI Overviews containing external links

Search Engine Roundtablesourcecitations sourcesevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-rate of a search with an AI Overview is reduced from 27% to 11% at position 1.

click-rate without AI Overview
27.0
click-rate with AI Overview
11.0

Compared metric: m1 — click-rate without AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

The click-through rate of a search with an AI Overview at position 1 is reduced from 27% to 11%.

click-rate at position 1 without AI Overview
27.0 percent
click-rate at position 1 with AI Overview
11.0 percent

Compared metric: m1 — click-rate at position 1 without AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-rate of a search with an AI Overview is reduced from 27% to 11% at position 1.

click-rate without AI Overview
27.0
click-rate with AI Overview
11.0

Compared metric: m1 — click-rate without AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

The click-through rate of a search with an AI Overview at position 1 is reduced from 27% to 11%.

click-rate at position 1 without AI Overview
27.0 percent
click-rate at position 1 with AI Overview
11.0 percent

Compared metric: m2 — click-rate at position 1 with AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-rate of a search with an AI Overview is reduced from 27% to 11% at position 1.

click-rate without AI Overview
27.0
click-rate with AI Overview
11.0

Compared metric: m1 — click-rate without AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

Google's AI Overviews now reaches 2.5 billion monthly users.

monthly users
2500000000.0 users · monthly

Compared metric: m1 — monthly users

SISTRIXsourceaudience usageevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-rate of a search with an AI Overview is reduced from 27% to 11% at position 1.

click-rate without AI Overview
27.0
click-rate with AI Overview
11.0

Compared metric: m1 — click-rate without AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

AI Overviews in France cut click-through rate (CTR) by 23.1%.

CTR cut by AI Overviews
23.1

Compared metric: m1 — CTR cut by AI Overviews

Ahrefssourceaudience usageevidence for google-ai-overviews

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-rate of a search with an AI Overview is reduced from 27% to 11% at position 1.

click-rate without AI Overview
27.0
click-rate with AI Overview
11.0

Compared metric: m1 — click-rate without AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

The click-through rate for a search result at position 1 is reduced from 27% to 11% when an AI Overview is present.

click-through rate
27.0 · position 1 search result without AI Overview
click-through rate
11.0 · position 1 search result with AI Overview

Compared metric: m2 — click-through rate

SISTRIXsourceoptimisation implicationevidence for google-ai-overviews

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-rate of a search with an AI Overview is reduced from 27% to 11% at position 1.

click-rate without AI Overview
27.0
click-rate with AI Overview
11.0

Compared metric: m1 — click-rate without AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

Google's AI Overviews reaches 2.5 billion monthly users.

monthly users
2500000000.0 · AI Overviews

Compared metric: m1 — monthly users

SISTRIXsourceaudience usageevidence for google-ai-overviews

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-rate of a search with an AI Overview is reduced from 27% to 11% at position 1.

click-rate without AI Overview
27.0
click-rate with AI Overview
11.0

Compared metric: m1 — click-rate without AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

Reddit's citation share also fell in Google's AI Overviews and AI Mode over the same period, by 11% and 31% respectively, which is shallower than the ChatGPT drop.

fall in citation share
11.0 · over the same period · in Google's AI Overviews
fall in citation share
31.0 · over the same period · in Google's AI Mode

Compared metric: m1 — fall in citation share

Semrushsourcecitations sourcesevidence for google-ai-overviews

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-rate of a search with an AI Overview is reduced from 27% to 11% at position 1.

click-rate without AI Overview
27.0
click-rate with AI Overview
11.0

Compared metric: m1 — click-rate without AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

Reddit's citation share also fell in Google's AI Overviews and AI Mode over the same period, by 11% and 31% respectively, which is shallower than the ChatGPT drop.

fall in citation share
11.0 · over the same period · in Google's AI Overviews
fall in citation share
31.0 · over the same period · in Google's AI Mode

Compared metric: m2 — fall in citation share

Semrushsourcecitations sourcesevidence for google-ai-overviews

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-rate of a search with an AI Overview is reduced from 27% to 11% at position 1.

click-rate without AI Overview
27.0
click-rate with AI Overview
11.0

Compared metric: m2 — click-rate with AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

Google's AI Overviews have significantly increased their inclusion of external links, rising from nearly 0% to over 25% of AI Overviews.

percentage of AI Overviews containing external links
0.0 % · AI Overviews
percentage of AI Overviews containing external links
25.0 % · AI Overviews

Compared metric: m1 — percentage of AI Overviews containing external links

Search Engine Roundtablesourcecitations sourcesevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-rate of a search with an AI Overview is reduced from 27% to 11% at position 1.

click-rate without AI Overview
27.0
click-rate with AI Overview
11.0

Compared metric: m2 — click-rate with AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

Google's AI Overviews have significantly increased their inclusion of external links, rising from nearly 0% to over 25% of AI Overviews.

percentage of AI Overviews containing external links
0.0 % · AI Overviews
percentage of AI Overviews containing external links
25.0 % · AI Overviews

Compared metric: m2 — percentage of AI Overviews containing external links

Search Engine Roundtablesourcecitations sourcesevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-rate of a search with an AI Overview is reduced from 27% to 11% at position 1.

click-rate without AI Overview
27.0
click-rate with AI Overview
11.0

Compared metric: m2 — click-rate with AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

The click-through rate of a search with an AI Overview at position 1 is reduced from 27% to 11%.

click-rate at position 1 without AI Overview
27.0 percent
click-rate at position 1 with AI Overview
11.0 percent

Compared metric: m1 — click-rate at position 1 without AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-rate of a search with an AI Overview is reduced from 27% to 11% at position 1.

click-rate without AI Overview
27.0
click-rate with AI Overview
11.0

Compared metric: m2 — click-rate with AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

The click-through rate of a search with an AI Overview at position 1 is reduced from 27% to 11%.

click-rate at position 1 without AI Overview
27.0 percent
click-rate at position 1 with AI Overview
11.0 percent

Compared metric: m2 — click-rate at position 1 with AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-rate of a search with an AI Overview is reduced from 27% to 11% at position 1.

click-rate without AI Overview
27.0
click-rate with AI Overview
11.0

Compared metric: m2 — click-rate with AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

Google's AI Overviews now reaches 2.5 billion monthly users.

monthly users
2500000000.0 users · monthly

Compared metric: m1 — monthly users

SISTRIXsourceaudience usageevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-rate of a search with an AI Overview is reduced from 27% to 11% at position 1.

click-rate without AI Overview
27.0
click-rate with AI Overview
11.0

Compared metric: m2 — click-rate with AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

AI Overviews in France cut click-through rate (CTR) by 23.1%.

CTR cut by AI Overviews
23.1

Compared metric: m1 — CTR cut by AI Overviews

Ahrefssourceaudience usageevidence for google-ai-overviews

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-rate of a search with an AI Overview is reduced from 27% to 11% at position 1.

click-rate without AI Overview
27.0
click-rate with AI Overview
11.0

Compared metric: m2 — click-rate with AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

The click-through rate for a search result at position 1 is reduced from 27% to 11% when an AI Overview is present.

click-through rate
27.0 · position 1 search result without AI Overview
click-through rate
11.0 · position 1 search result with AI Overview

Compared metric: m1 — click-through rate

SISTRIXsourceoptimisation implicationevidence for google-ai-overviews

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-rate of a search with an AI Overview is reduced from 27% to 11% at position 1.

click-rate without AI Overview
27.0
click-rate with AI Overview
11.0

Compared metric: m2 — click-rate with AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

Google's AI Overviews reaches 2.5 billion monthly users.

monthly users
2500000000.0 · AI Overviews

Compared metric: m1 — monthly users

SISTRIXsourceaudience usageevidence for google-ai-overviews

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-rate of a search with an AI Overview is reduced from 27% to 11% at position 1.

click-rate without AI Overview
27.0
click-rate with AI Overview
11.0

Compared metric: m2 — click-rate with AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

Reddit's citation share also fell in Google's AI Overviews and AI Mode over the same period, by 11% and 31% respectively, which is shallower than the ChatGPT drop.

fall in citation share
11.0 · over the same period · in Google's AI Overviews
fall in citation share
31.0 · over the same period · in Google's AI Mode

Compared metric: m2 — fall in citation share

Semrushsourcecitations sourcesevidence for google-ai-overviews

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google's AI Overviews have significantly increased their inclusion of external links, rising from nearly 0% to over 25% of AI Overviews.

percentage of AI Overviews containing external links
0.0 % · AI Overviews
percentage of AI Overviews containing external links
25.0 % · AI Overviews

Compared metric: m1 — percentage of AI Overviews containing external links

Search Engine Roundtablesourcecitations sourcesevidence for google-ai-overviews

The click-through rate of a search with an AI Overview at position 1 is reduced from 27% to 11%.

click-rate at position 1 without AI Overview
27.0 percent
click-rate at position 1 with AI Overview
11.0 percent

Compared metric: m1 — click-rate at position 1 without AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google's AI Overviews have significantly increased their inclusion of external links, rising from nearly 0% to over 25% of AI Overviews.

percentage of AI Overviews containing external links
0.0 % · AI Overviews
percentage of AI Overviews containing external links
25.0 % · AI Overviews

Compared metric: m1 — percentage of AI Overviews containing external links

Search Engine Roundtablesourcecitations sourcesevidence for google-ai-overviews

The click-through rate of a search with an AI Overview at position 1 is reduced from 27% to 11%.

click-rate at position 1 without AI Overview
27.0 percent
click-rate at position 1 with AI Overview
11.0 percent

Compared metric: m2 — click-rate at position 1 with AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google's AI Overviews have significantly increased their inclusion of external links, rising from nearly 0% to over 25% of AI Overviews.

percentage of AI Overviews containing external links
0.0 % · AI Overviews
percentage of AI Overviews containing external links
25.0 % · AI Overviews

Compared metric: m1 — percentage of AI Overviews containing external links

Search Engine Roundtablesourcecitations sourcesevidence for google-ai-overviews

Google's AI Overviews now reaches 2.5 billion monthly users.

monthly users
2500000000.0 users · monthly

Compared metric: m1 — monthly users

SISTRIXsourceaudience usageevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google's AI Overviews have significantly increased their inclusion of external links, rising from nearly 0% to over 25% of AI Overviews.

percentage of AI Overviews containing external links
0.0 % · AI Overviews
percentage of AI Overviews containing external links
25.0 % · AI Overviews

Compared metric: m1 — percentage of AI Overviews containing external links

Search Engine Roundtablesourcecitations sourcesevidence for google-ai-overviews

AI Overviews in France cut click-through rate (CTR) by 23.1%.

CTR cut by AI Overviews
23.1

Compared metric: m1 — CTR cut by AI Overviews

Ahrefssourceaudience usageevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google's AI Overviews have significantly increased their inclusion of external links, rising from nearly 0% to over 25% of AI Overviews.

percentage of AI Overviews containing external links
0.0 % · AI Overviews
percentage of AI Overviews containing external links
25.0 % · AI Overviews

Compared metric: m1 — percentage of AI Overviews containing external links

Search Engine Roundtablesourcecitations sourcesevidence for google-ai-overviews

The click-through rate for a search result at position 1 is reduced from 27% to 11% when an AI Overview is present.

click-through rate
27.0 · position 1 search result without AI Overview
click-through rate
11.0 · position 1 search result with AI Overview

Compared metric: m1 — click-through rate

SISTRIXsourceoptimisation implicationevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google's AI Overviews have significantly increased their inclusion of external links, rising from nearly 0% to over 25% of AI Overviews.

percentage of AI Overviews containing external links
0.0 % · AI Overviews
percentage of AI Overviews containing external links
25.0 % · AI Overviews

Compared metric: m1 — percentage of AI Overviews containing external links

Search Engine Roundtablesourcecitations sourcesevidence for google-ai-overviews

The click-through rate for a search result at position 1 is reduced from 27% to 11% when an AI Overview is present.

click-through rate
27.0 · position 1 search result without AI Overview
click-through rate
11.0 · position 1 search result with AI Overview

Compared metric: m2 — click-through rate

SISTRIXsourceoptimisation implicationevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google's AI Overviews have significantly increased their inclusion of external links, rising from nearly 0% to over 25% of AI Overviews.

percentage of AI Overviews containing external links
0.0 % · AI Overviews
percentage of AI Overviews containing external links
25.0 % · AI Overviews

Compared metric: m1 — percentage of AI Overviews containing external links

Search Engine Roundtablesourcecitations sourcesevidence for google-ai-overviews

Google's AI Overviews reaches 2.5 billion monthly users.

monthly users
2500000000.0 · AI Overviews

Compared metric: m1 — monthly users

SISTRIXsourceaudience usageevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google's AI Overviews have significantly increased their inclusion of external links, rising from nearly 0% to over 25% of AI Overviews.

percentage of AI Overviews containing external links
0.0 % · AI Overviews
percentage of AI Overviews containing external links
25.0 % · AI Overviews

Compared metric: m1 — percentage of AI Overviews containing external links

Search Engine Roundtablesourcecitations sourcesevidence for google-ai-overviews

Reddit's citation share also fell in Google's AI Overviews and AI Mode over the same period, by 11% and 31% respectively, which is shallower than the ChatGPT drop.

fall in citation share
11.0 · over the same period · in Google's AI Overviews
fall in citation share
31.0 · over the same period · in Google's AI Mode

Compared metric: m1 — fall in citation share

Semrushsourcecitations sourcesevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google's AI Overviews have significantly increased their inclusion of external links, rising from nearly 0% to over 25% of AI Overviews.

percentage of AI Overviews containing external links
0.0 % · AI Overviews
percentage of AI Overviews containing external links
25.0 % · AI Overviews

Compared metric: m1 — percentage of AI Overviews containing external links

Search Engine Roundtablesourcecitations sourcesevidence for google-ai-overviews

Reddit's citation share also fell in Google's AI Overviews and AI Mode over the same period, by 11% and 31% respectively, which is shallower than the ChatGPT drop.

fall in citation share
11.0 · over the same period · in Google's AI Overviews
fall in citation share
31.0 · over the same period · in Google's AI Mode

Compared metric: m2 — fall in citation share

Semrushsourcecitations sourcesevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google's AI Overviews have significantly increased their inclusion of external links, rising from nearly 0% to over 25% of AI Overviews.

percentage of AI Overviews containing external links
0.0 % · AI Overviews
percentage of AI Overviews containing external links
25.0 % · AI Overviews

Compared metric: m2 — percentage of AI Overviews containing external links

Search Engine Roundtablesourcecitations sourcesevidence for google-ai-overviews

The click-through rate of a search with an AI Overview at position 1 is reduced from 27% to 11%.

click-rate at position 1 without AI Overview
27.0 percent
click-rate at position 1 with AI Overview
11.0 percent

Compared metric: m1 — click-rate at position 1 without AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google's AI Overviews have significantly increased their inclusion of external links, rising from nearly 0% to over 25% of AI Overviews.

percentage of AI Overviews containing external links
0.0 % · AI Overviews
percentage of AI Overviews containing external links
25.0 % · AI Overviews

Compared metric: m2 — percentage of AI Overviews containing external links

Search Engine Roundtablesourcecitations sourcesevidence for google-ai-overviews

The click-through rate of a search with an AI Overview at position 1 is reduced from 27% to 11%.

click-rate at position 1 without AI Overview
27.0 percent
click-rate at position 1 with AI Overview
11.0 percent

Compared metric: m2 — click-rate at position 1 with AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google's AI Overviews have significantly increased their inclusion of external links, rising from nearly 0% to over 25% of AI Overviews.

percentage of AI Overviews containing external links
0.0 % · AI Overviews
percentage of AI Overviews containing external links
25.0 % · AI Overviews

Compared metric: m2 — percentage of AI Overviews containing external links

Search Engine Roundtablesourcecitations sourcesevidence for google-ai-overviews

Google's AI Overviews now reaches 2.5 billion monthly users.

monthly users
2500000000.0 users · monthly

Compared metric: m1 — monthly users

SISTRIXsourceaudience usageevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google's AI Overviews have significantly increased their inclusion of external links, rising from nearly 0% to over 25% of AI Overviews.

percentage of AI Overviews containing external links
0.0 % · AI Overviews
percentage of AI Overviews containing external links
25.0 % · AI Overviews

Compared metric: m2 — percentage of AI Overviews containing external links

Search Engine Roundtablesourcecitations sourcesevidence for google-ai-overviews

AI Overviews in France cut click-through rate (CTR) by 23.1%.

CTR cut by AI Overviews
23.1

Compared metric: m1 — CTR cut by AI Overviews

Ahrefssourceaudience usageevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google's AI Overviews have significantly increased their inclusion of external links, rising from nearly 0% to over 25% of AI Overviews.

percentage of AI Overviews containing external links
0.0 % · AI Overviews
percentage of AI Overviews containing external links
25.0 % · AI Overviews

Compared metric: m2 — percentage of AI Overviews containing external links

Search Engine Roundtablesourcecitations sourcesevidence for google-ai-overviews

The click-through rate for a search result at position 1 is reduced from 27% to 11% when an AI Overview is present.

click-through rate
27.0 · position 1 search result without AI Overview
click-through rate
11.0 · position 1 search result with AI Overview

Compared metric: m1 — click-through rate

SISTRIXsourceoptimisation implicationevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google's AI Overviews have significantly increased their inclusion of external links, rising from nearly 0% to over 25% of AI Overviews.

percentage of AI Overviews containing external links
0.0 % · AI Overviews
percentage of AI Overviews containing external links
25.0 % · AI Overviews

Compared metric: m2 — percentage of AI Overviews containing external links

Search Engine Roundtablesourcecitations sourcesevidence for google-ai-overviews

The click-through rate for a search result at position 1 is reduced from 27% to 11% when an AI Overview is present.

click-through rate
27.0 · position 1 search result without AI Overview
click-through rate
11.0 · position 1 search result with AI Overview

Compared metric: m2 — click-through rate

SISTRIXsourceoptimisation implicationevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google's AI Overviews have significantly increased their inclusion of external links, rising from nearly 0% to over 25% of AI Overviews.

percentage of AI Overviews containing external links
0.0 % · AI Overviews
percentage of AI Overviews containing external links
25.0 % · AI Overviews

Compared metric: m2 — percentage of AI Overviews containing external links

Search Engine Roundtablesourcecitations sourcesevidence for google-ai-overviews

Google's AI Overviews reaches 2.5 billion monthly users.

monthly users
2500000000.0 · AI Overviews

Compared metric: m1 — monthly users

SISTRIXsourceaudience usageevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google's AI Overviews have significantly increased their inclusion of external links, rising from nearly 0% to over 25% of AI Overviews.

percentage of AI Overviews containing external links
0.0 % · AI Overviews
percentage of AI Overviews containing external links
25.0 % · AI Overviews

Compared metric: m2 — percentage of AI Overviews containing external links

Search Engine Roundtablesourcecitations sourcesevidence for google-ai-overviews

Reddit's citation share also fell in Google's AI Overviews and AI Mode over the same period, by 11% and 31% respectively, which is shallower than the ChatGPT drop.

fall in citation share
11.0 · over the same period · in Google's AI Overviews
fall in citation share
31.0 · over the same period · in Google's AI Mode

Compared metric: m1 — fall in citation share

Semrushsourcecitations sourcesevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google's AI Overviews have significantly increased their inclusion of external links, rising from nearly 0% to over 25% of AI Overviews.

percentage of AI Overviews containing external links
0.0 % · AI Overviews
percentage of AI Overviews containing external links
25.0 % · AI Overviews

Compared metric: m2 — percentage of AI Overviews containing external links

Search Engine Roundtablesourcecitations sourcesevidence for google-ai-overviews

Reddit's citation share also fell in Google's AI Overviews and AI Mode over the same period, by 11% and 31% respectively, which is shallower than the ChatGPT drop.

fall in citation share
11.0 · over the same period · in Google's AI Overviews
fall in citation share
31.0 · over the same period · in Google's AI Mode

Compared metric: m2 — fall in citation share

Semrushsourcecitations sourcesevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-through rate of a search with an AI Overview at position 1 is reduced from 27% to 11%.

click-rate at position 1 without AI Overview
27.0 percent
click-rate at position 1 with AI Overview
11.0 percent

Compared metric: m1 — click-rate at position 1 without AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

The click-through rate of a search with an AI Overview at position 1 is reduced from 27% to 11%.

click-rate at position 1 without AI Overview
27.0 percent
click-rate at position 1 with AI Overview
11.0 percent

Compared metric: m2 — click-rate at position 1 with AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-through rate of a search with an AI Overview at position 1 is reduced from 27% to 11%.

click-rate at position 1 without AI Overview
27.0 percent
click-rate at position 1 with AI Overview
11.0 percent

Compared metric: m1 — click-rate at position 1 without AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

Google's AI Overviews now reaches 2.5 billion monthly users.

monthly users
2500000000.0 users · monthly

Compared metric: m1 — monthly users

SISTRIXsourceaudience usageevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-through rate of a search with an AI Overview at position 1 is reduced from 27% to 11%.

click-rate at position 1 without AI Overview
27.0 percent
click-rate at position 1 with AI Overview
11.0 percent

Compared metric: m1 — click-rate at position 1 without AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

AI Overviews in France cut click-through rate (CTR) by 23.1%.

CTR cut by AI Overviews
23.1

Compared metric: m1 — CTR cut by AI Overviews

Ahrefssourceaudience usageevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-through rate of a search with an AI Overview at position 1 is reduced from 27% to 11%.

click-rate at position 1 without AI Overview
27.0 percent
click-rate at position 1 with AI Overview
11.0 percent

Compared metric: m1 — click-rate at position 1 without AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

The click-through rate for a search result at position 1 is reduced from 27% to 11% when an AI Overview is present.

click-through rate
27.0 · position 1 search result without AI Overview
click-through rate
11.0 · position 1 search result with AI Overview

Compared metric: m1 — click-through rate

SISTRIXsourceoptimisation implicationevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-through rate of a search with an AI Overview at position 1 is reduced from 27% to 11%.

click-rate at position 1 without AI Overview
27.0 percent
click-rate at position 1 with AI Overview
11.0 percent

Compared metric: m1 — click-rate at position 1 without AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

The click-through rate for a search result at position 1 is reduced from 27% to 11% when an AI Overview is present.

click-through rate
27.0 · position 1 search result without AI Overview
click-through rate
11.0 · position 1 search result with AI Overview

Compared metric: m2 — click-through rate

SISTRIXsourceoptimisation implicationevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-through rate of a search with an AI Overview at position 1 is reduced from 27% to 11%.

click-rate at position 1 without AI Overview
27.0 percent
click-rate at position 1 with AI Overview
11.0 percent

Compared metric: m1 — click-rate at position 1 without AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

Google's AI Overviews reaches 2.5 billion monthly users.

monthly users
2500000000.0 · AI Overviews

Compared metric: m1 — monthly users

SISTRIXsourceaudience usageevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-through rate of a search with an AI Overview at position 1 is reduced from 27% to 11%.

click-rate at position 1 without AI Overview
27.0 percent
click-rate at position 1 with AI Overview
11.0 percent

Compared metric: m1 — click-rate at position 1 without AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

Reddit's citation share also fell in Google's AI Overviews and AI Mode over the same period, by 11% and 31% respectively, which is shallower than the ChatGPT drop.

fall in citation share
11.0 · over the same period · in Google's AI Overviews
fall in citation share
31.0 · over the same period · in Google's AI Mode

Compared metric: m1 — fall in citation share

Semrushsourcecitations sourcesevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-through rate of a search with an AI Overview at position 1 is reduced from 27% to 11%.

click-rate at position 1 without AI Overview
27.0 percent
click-rate at position 1 with AI Overview
11.0 percent

Compared metric: m1 — click-rate at position 1 without AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

Reddit's citation share also fell in Google's AI Overviews and AI Mode over the same period, by 11% and 31% respectively, which is shallower than the ChatGPT drop.

fall in citation share
11.0 · over the same period · in Google's AI Overviews
fall in citation share
31.0 · over the same period · in Google's AI Mode

Compared metric: m2 — fall in citation share

Semrushsourcecitations sourcesevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-through rate of a search with an AI Overview at position 1 is reduced from 27% to 11%.

click-rate at position 1 without AI Overview
27.0 percent
click-rate at position 1 with AI Overview
11.0 percent

Compared metric: m2 — click-rate at position 1 with AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

Google's AI Overviews now reaches 2.5 billion monthly users.

monthly users
2500000000.0 users · monthly

Compared metric: m1 — monthly users

SISTRIXsourceaudience usageevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-through rate of a search with an AI Overview at position 1 is reduced from 27% to 11%.

click-rate at position 1 without AI Overview
27.0 percent
click-rate at position 1 with AI Overview
11.0 percent

Compared metric: m2 — click-rate at position 1 with AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

AI Overviews in France cut click-through rate (CTR) by 23.1%.

CTR cut by AI Overviews
23.1

Compared metric: m1 — CTR cut by AI Overviews

Ahrefssourceaudience usageevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-through rate of a search with an AI Overview at position 1 is reduced from 27% to 11%.

click-rate at position 1 without AI Overview
27.0 percent
click-rate at position 1 with AI Overview
11.0 percent

Compared metric: m2 — click-rate at position 1 with AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

The click-through rate for a search result at position 1 is reduced from 27% to 11% when an AI Overview is present.

click-through rate
27.0 · position 1 search result without AI Overview
click-through rate
11.0 · position 1 search result with AI Overview

Compared metric: m1 — click-through rate

SISTRIXsourceoptimisation implicationevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-through rate of a search with an AI Overview at position 1 is reduced from 27% to 11%.

click-rate at position 1 without AI Overview
27.0 percent
click-rate at position 1 with AI Overview
11.0 percent

Compared metric: m2 — click-rate at position 1 with AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

The click-through rate for a search result at position 1 is reduced from 27% to 11% when an AI Overview is present.

click-through rate
27.0 · position 1 search result without AI Overview
click-through rate
11.0 · position 1 search result with AI Overview

Compared metric: m2 — click-through rate

SISTRIXsourceoptimisation implicationevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-through rate of a search with an AI Overview at position 1 is reduced from 27% to 11%.

click-rate at position 1 without AI Overview
27.0 percent
click-rate at position 1 with AI Overview
11.0 percent

Compared metric: m2 — click-rate at position 1 with AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

Google's AI Overviews reaches 2.5 billion monthly users.

monthly users
2500000000.0 · AI Overviews

Compared metric: m1 — monthly users

SISTRIXsourceaudience usageevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-through rate of a search with an AI Overview at position 1 is reduced from 27% to 11%.

click-rate at position 1 without AI Overview
27.0 percent
click-rate at position 1 with AI Overview
11.0 percent

Compared metric: m2 — click-rate at position 1 with AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

Reddit's citation share also fell in Google's AI Overviews and AI Mode over the same period, by 11% and 31% respectively, which is shallower than the ChatGPT drop.

fall in citation share
11.0 · over the same period · in Google's AI Overviews
fall in citation share
31.0 · over the same period · in Google's AI Mode

Compared metric: m1 — fall in citation share

Semrushsourcecitations sourcesevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-through rate of a search with an AI Overview at position 1 is reduced from 27% to 11%.

click-rate at position 1 without AI Overview
27.0 percent
click-rate at position 1 with AI Overview
11.0 percent

Compared metric: m2 — click-rate at position 1 with AI Overview

SISTRIXsourcemeasurementevidence for google-ai-overviews

Reddit's citation share also fell in Google's AI Overviews and AI Mode over the same period, by 11% and 31% respectively, which is shallower than the ChatGPT drop.

fall in citation share
11.0 · over the same period · in Google's AI Overviews
fall in citation share
31.0 · over the same period · in Google's AI Mode

Compared metric: m2 — fall in citation share

Semrushsourcecitations sourcesevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google's AI Overviews now reaches 2.5 billion monthly users.

monthly users
2500000000.0 users · monthly

Compared metric: m1 — monthly users

SISTRIXsourceaudience usageevidence for google-ai-overviews

AI Overviews in France cut click-through rate (CTR) by 23.1%.

CTR cut by AI Overviews
23.1

Compared metric: m1 — CTR cut by AI Overviews

Ahrefssourceaudience usageevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google's AI Overviews now reaches 2.5 billion monthly users.

monthly users
2500000000.0 users · monthly

Compared metric: m1 — monthly users

SISTRIXsourceaudience usageevidence for google-ai-overviews

The click-through rate for a search result at position 1 is reduced from 27% to 11% when an AI Overview is present.

click-through rate
27.0 · position 1 search result without AI Overview
click-through rate
11.0 · position 1 search result with AI Overview

Compared metric: m1 — click-through rate

SISTRIXsourceoptimisation implicationevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google's AI Overviews now reaches 2.5 billion monthly users.

monthly users
2500000000.0 users · monthly

Compared metric: m1 — monthly users

SISTRIXsourceaudience usageevidence for google-ai-overviews

The click-through rate for a search result at position 1 is reduced from 27% to 11% when an AI Overview is present.

click-through rate
27.0 · position 1 search result without AI Overview
click-through rate
11.0 · position 1 search result with AI Overview

Compared metric: m2 — click-through rate

SISTRIXsourceoptimisation implicationevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google's AI Overviews now reaches 2.5 billion monthly users.

monthly users
2500000000.0 users · monthly

Compared metric: m1 — monthly users

SISTRIXsourceaudience usageevidence for google-ai-overviews

Google's AI Overviews reaches 2.5 billion monthly users.

monthly users
2500000000.0 · AI Overviews

Compared metric: m1 — monthly users

SISTRIXsourceaudience usageevidence for google-ai-overviews

Agency interpretationThese findings differ in unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google's AI Overviews now reaches 2.5 billion monthly users.

monthly users
2500000000.0 users · monthly

Compared metric: m1 — monthly users

SISTRIXsourceaudience usageevidence for google-ai-overviews

Reddit's citation share also fell in Google's AI Overviews and AI Mode over the same period, by 11% and 31% respectively, which is shallower than the ChatGPT drop.

fall in citation share
11.0 · over the same period · in Google's AI Overviews
fall in citation share
31.0 · over the same period · in Google's AI Mode

Compared metric: m1 — fall in citation share

Semrushsourcecitations sourcesevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google's AI Overviews now reaches 2.5 billion monthly users.

monthly users
2500000000.0 users · monthly

Compared metric: m1 — monthly users

SISTRIXsourceaudience usageevidence for google-ai-overviews

Reddit's citation share also fell in Google's AI Overviews and AI Mode over the same period, by 11% and 31% respectively, which is shallower than the ChatGPT drop.

fall in citation share
11.0 · over the same period · in Google's AI Overviews
fall in citation share
31.0 · over the same period · in Google's AI Mode

Compared metric: m2 — fall in citation share

Semrushsourcecitations sourcesevidence for google-ai-overviews

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI Overviews in France cut click-through rate (CTR) by 23.1%.

CTR cut by AI Overviews
23.1

Compared metric: m1 — CTR cut by AI Overviews

Ahrefssourceaudience usageevidence for google-ai-overviews

The click-through rate for a search result at position 1 is reduced from 27% to 11% when an AI Overview is present.

click-through rate
27.0 · position 1 search result without AI Overview
click-through rate
11.0 · position 1 search result with AI Overview

Compared metric: m1 — click-through rate

SISTRIXsourceoptimisation implicationevidence for google-ai-overviews

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI Overviews in France cut click-through rate (CTR) by 23.1%.

CTR cut by AI Overviews
23.1

Compared metric: m1 — CTR cut by AI Overviews

Ahrefssourceaudience usageevidence for google-ai-overviews

The click-through rate for a search result at position 1 is reduced from 27% to 11% when an AI Overview is present.

click-through rate
27.0 · position 1 search result without AI Overview
click-through rate
11.0 · position 1 search result with AI Overview

Compared metric: m2 — click-through rate

SISTRIXsourceoptimisation implicationevidence for google-ai-overviews

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI Overviews in France cut click-through rate (CTR) by 23.1%.

CTR cut by AI Overviews
23.1

Compared metric: m1 — CTR cut by AI Overviews

Ahrefssourceaudience usageevidence for google-ai-overviews

Google's AI Overviews reaches 2.5 billion monthly users.

monthly users
2500000000.0 · AI Overviews

Compared metric: m1 — monthly users

SISTRIXsourceaudience usageevidence for google-ai-overviews

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI Overviews in France cut click-through rate (CTR) by 23.1%.

CTR cut by AI Overviews
23.1

Compared metric: m1 — CTR cut by AI Overviews

Ahrefssourceaudience usageevidence for google-ai-overviews

Reddit's citation share also fell in Google's AI Overviews and AI Mode over the same period, by 11% and 31% respectively, which is shallower than the ChatGPT drop.

fall in citation share
11.0 · over the same period · in Google's AI Overviews
fall in citation share
31.0 · over the same period · in Google's AI Mode

Compared metric: m1 — fall in citation share

Semrushsourcecitations sourcesevidence for google-ai-overviews

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI Overviews in France cut click-through rate (CTR) by 23.1%.

CTR cut by AI Overviews
23.1

Compared metric: m1 — CTR cut by AI Overviews

Ahrefssourceaudience usageevidence for google-ai-overviews

Reddit's citation share also fell in Google's AI Overviews and AI Mode over the same period, by 11% and 31% respectively, which is shallower than the ChatGPT drop.

fall in citation share
11.0 · over the same period · in Google's AI Overviews
fall in citation share
31.0 · over the same period · in Google's AI Mode

Compared metric: m2 — fall in citation share

Semrushsourcecitations sourcesevidence for google-ai-overviews

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-through rate for a search result at position 1 is reduced from 27% to 11% when an AI Overview is present.

click-through rate
27.0 · position 1 search result without AI Overview
click-through rate
11.0 · position 1 search result with AI Overview

Compared metric: m1 — click-through rate

SISTRIXsourceoptimisation implicationevidence for google-ai-overviews

Google's AI Overviews reaches 2.5 billion monthly users.

monthly users
2500000000.0 · AI Overviews

Compared metric: m1 — monthly users

SISTRIXsourceaudience usageevidence for google-ai-overviews

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-through rate for a search result at position 1 is reduced from 27% to 11% when an AI Overview is present.

click-through rate
27.0 · position 1 search result without AI Overview
click-through rate
11.0 · position 1 search result with AI Overview

Compared metric: m1 — click-through rate

SISTRIXsourceoptimisation implicationevidence for google-ai-overviews

Reddit's citation share also fell in Google's AI Overviews and AI Mode over the same period, by 11% and 31% respectively, which is shallower than the ChatGPT drop.

fall in citation share
11.0 · over the same period · in Google's AI Overviews
fall in citation share
31.0 · over the same period · in Google's AI Mode

Compared metric: m1 — fall in citation share

Semrushsourcecitations sourcesevidence for google-ai-overviews

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-through rate for a search result at position 1 is reduced from 27% to 11% when an AI Overview is present.

click-through rate
27.0 · position 1 search result without AI Overview
click-through rate
11.0 · position 1 search result with AI Overview

Compared metric: m1 — click-through rate

SISTRIXsourceoptimisation implicationevidence for google-ai-overviews

Reddit's citation share also fell in Google's AI Overviews and AI Mode over the same period, by 11% and 31% respectively, which is shallower than the ChatGPT drop.

fall in citation share
11.0 · over the same period · in Google's AI Overviews
fall in citation share
31.0 · over the same period · in Google's AI Mode

Compared metric: m2 — fall in citation share

Semrushsourcecitations sourcesevidence for google-ai-overviews

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-through rate for a search result at position 1 is reduced from 27% to 11% when an AI Overview is present.

click-through rate
27.0 · position 1 search result without AI Overview
click-through rate
11.0 · position 1 search result with AI Overview

Compared metric: m2 — click-through rate

SISTRIXsourceoptimisation implicationevidence for google-ai-overviews

Google's AI Overviews reaches 2.5 billion monthly users.

monthly users
2500000000.0 · AI Overviews

Compared metric: m1 — monthly users

SISTRIXsourceaudience usageevidence for google-ai-overviews

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The click-through rate for a search result at position 1 is reduced from 27% to 11% when an AI Overview is present.

click-through rate
27.0 · position 1 search result without AI Overview
click-through rate
11.0 · position 1 search result with AI Overview

Compared metric: m2 — click-through rate

SISTRIXsourceoptimisation implicationevidence for google-ai-overviews

Reddit's citation share also fell in Google's AI Overviews and AI Mode over the same period, by 11% and 31% respectively, which is shallower than the ChatGPT drop.

fall in citation share
11.0 · over the same period · in Google's AI Overviews
fall in citation share
31.0 · over the same period · in Google's AI Mode

Compared metric: m2 — fall in citation share

Semrushsourcecitations sourcesevidence for google-ai-overviews

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google's AI Overviews reaches 2.5 billion monthly users.

monthly users
2500000000.0 · AI Overviews

Compared metric: m1 — monthly users

SISTRIXsourceaudience usageevidence for google-ai-overviews

Reddit's citation share also fell in Google's AI Overviews and AI Mode over the same period, by 11% and 31% respectively, which is shallower than the ChatGPT drop.

fall in citation share
11.0 · over the same period · in Google's AI Overviews
fall in citation share
31.0 · over the same period · in Google's AI Mode

Compared metric: m1 — fall in citation share

Semrushsourcecitations sourcesevidence for google-ai-overviews

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google's AI Overviews reaches 2.5 billion monthly users.

monthly users
2500000000.0 · AI Overviews

Compared metric: m1 — monthly users

SISTRIXsourceaudience usageevidence for google-ai-overviews

Reddit's citation share also fell in Google's AI Overviews and AI Mode over the same period, by 11% and 31% respectively, which is shallower than the ChatGPT drop.

fall in citation share
11.0 · over the same period · in Google's AI Overviews
fall in citation share
31.0 · over the same period · in Google's AI Mode

Compared metric: m2 — fall in citation share

Semrushsourcecitations sourcesevidence for google-ai-overviews

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A typical Google Core Update shifts the share of top domains by 5–10% when measured by the visibility index, whereas the GPT-5.5 rollout caused a 47% citation drift.

citation drift
47.0
shift in top domains for Google Core Upd
5–10

Compared metric: m1 — citation drift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The switch from GPT-5 mini to GPT-5.5 on May 23, 2026, shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0

Compared metric: m1 — citation shift

SISTRIXsourceoptimisation implicationevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying.

percentage of AI users talked out of buy
57.5

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourcereferrals conversionevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.

proportion of AI users who dislike chatb
4.0

Compared metric: m1 — proportion of AI users who dislike chatb

Semrushsourcecommerce adsevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.

percentage of AI users who have replaced
65.0

Compared metric: m1 — percentage of AI users who have replaced

Semrushsourceaudience usageevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m1 — citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 % · within 48 hours · German-language ChatGPT responses

Compared metric: m1 — citation shift

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

AI chatbots talked 57.5% of AI users out of buying, based on a survey.

percentage of AI users talked out of buy
57.5 percent · AI users

Compared metric: m1 — percentage of AI users talked out of buy

Semrushsourceaudience usageevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m1 — percentage of AI users who replaced prod

Semrushsourceaudience usageevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

65% of AI users have replaced some product-related Google searches with chatbots, and 4 in 10 dislike chatbot ads, according to a survey of 2,338 US consumers.

percentage of AI users who replaced prod
65.0 percent · AI users
percentage of AI users who dislike chatb
40.0 percent · AI users

Compared metric: m2 — percentage of AI users who dislike chatb

Semrushsourceaudience usageevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit saw a 59% increase in citations, gaining 7,007 citations per 10,000 responses, making it the biggest winner in the GPT-5.5 rollout in Germany.

citation increase
59.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses
citation increase
7007.0 citations per 10,000 responses · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m2 — citation increase

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

On May 23, 2026, the switch from GPT-5 mini to GPT-5.5 shifted citations in German-language ChatGPT responses by 47% within 48 hours.

citation shift
47.0 percent · within 48 hours

Compared metric: m1 — citation shift

SISTRIXsourceretrieval indexevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

ChatGPT held 79.4% of the AI chatbot market share worldwide in August 2026.

market share
79.4 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Initial observations of ChatGPT ads were based on three days of data when first published on July 15, and updated on July 28, 2026.

duration of ChatGPT ads data
3.0 days · ChatGPT ads data

Compared metric: m1 — duration of ChatGPT ads data

Seer Interactivesourcecommerce adsevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m1 — increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit's citations in German-language ChatGPT responses increased by 59% (+7,007 citations per 10,000 responses) after the GPT-5.5 rollout.

increase in citations
59.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses
absolute increase in citations
7007.0 · after GPT-5.5 rollout · Reddit in German-language ChatGPT responses

Compared metric: m2 — absolute increase in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

The domain distribution in German-language ChatGPT responses shifted by 47% within 48 hours during the GPT-5.5 rollout on May 22/23, 2026.

shift in domain distribution
47.0 · Within 48 hours · German-language ChatGPT responses

Compared metric: m1 — shift in domain distribution

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Reddit.com is the most cited domain by ChatGPT for US queries, accounting for 16.8% of all citations among top sources in September 2026.

mention share
16.8 percent · September 2026 · among the top sources

Compared metric: m1 — mention share

Ahrefssourcecitations sourcesevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m1 — average share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Between August 14 and 17, 2023, Reddit's share of ChatGPT citations averaged just 0.5%, an 86% decline relative to the previous period.

average share of citations
0.5 · Between August 14 and 17 · of ChatGPT citations
decline in share
86.0 · Between August 14 and 17 · relative to the previous period

Compared metric: m2 — decline in share

Semrushsourcecitations sourcesevidence for chatgpt

Reddit held a steady 3.8% share of ChatGPT citations from July 18 through August 7, 2023, which was one of the largest shares of any single domain.

share of citations
3.8 · from July 18 through August 7 · of ChatGPT citations

Compared metric: m1 — share of citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

More than half of consumers read summaries in Search, and those consumers are over 40% more likely to end their search after reading one.

proportion of consumers who read summari
0.5 percent
increased likelihood of consumers ending
40.0 percent

Compared metric: m1 — proportion of consumers who read summari

Cloudflaresourceaudience usageevidence for google-search-generative-experience

More than half of consumers read summaries in Search, and those consumers are over 40% more likely to end their search after reading one.

proportion of consumers who read summari
0.5 percent
increased likelihood of consumers ending
40.0 percent

Compared metric: m2 — increased likelihood of consumers ending

Cloudflaresourceaudience usageevidence for google-search-generative-experience

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

As of April 2026, DeepSeek's average monthly usage time per person was 109.5 minutes, a year-on-year increase of 106.9%.

average monthly usage time per person
109.5 minutes · 截止到2026年4月 · DeepSeek
year-on-year growth in average monthly usage time per person
106.9 % · 截止到2026年4月 · DeepSeek

Compared metric: m1 — average monthly usage time per person

QuestMobilesourceaudience usageevidence for deepseek

As of April 2026, DeepSeek's average monthly usage time per person was 109.5 minutes, a year-on-year increase of 106.9%.

average monthly usage time per person
109.5 minutes · 截止到2026年4月 · DeepSeek
year-on-year growth in average monthly usage time per person
106.9 % · 截止到2026年4月 · DeepSeek

Compared metric: m2 — year-on-year growth in average monthly usage time per person

QuestMobilesourceaudience usageevidence for deepseek

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

As of April 2026, DeepSeek's average monthly usage time per person was 109.5 minutes, a year-on-year increase of 106.9%.

average monthly usage time per person
109.5 minutes · 截止到2026年4月 · DeepSeek
year-on-year growth in average monthly usage time per person
106.9 % · 截止到2026年4月 · DeepSeek

Compared metric: m1 — average monthly usage time per person

QuestMobilesourceaudience usageevidence for deepseek

Deepseek held 0.02% of the AI chatbot market share worldwide in August 2026.

market share
0.02 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for deepseek

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

As of April 2026, DeepSeek's average monthly usage time per person was 109.5 minutes, a year-on-year increase of 106.9%.

average monthly usage time per person
109.5 minutes · 截止到2026年4月 · DeepSeek
year-on-year growth in average monthly usage time per person
106.9 % · 截止到2026年4月 · DeepSeek

Compared metric: m1 — average monthly usage time per person

QuestMobilesourceaudience usageevidence for deepseek

As of April 2026, DeepSeek's average monthly usage time per person was 109.5 minutes, a 106.9% year-on-year increase.

average monthly usage time per person
109.5 · 截止到2026年4月 · DeepSeek
year-on-year growth
106.9 · 截止到2026年4月 · DeepSeek

Compared metric: m1 — average monthly usage time per person

QuestMobilesourceaudience usageevidence for deepseek

Agency interpretationThese findings differ in unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • unit

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

As of April 2026, DeepSeek's average monthly usage time per person was 109.5 minutes, a year-on-year increase of 106.9%.

average monthly usage time per person
109.5 minutes · 截止到2026年4月 · DeepSeek
year-on-year growth in average monthly usage time per person
106.9 % · 截止到2026年4月 · DeepSeek

Compared metric: m1 — average monthly usage time per person

QuestMobilesourceaudience usageevidence for deepseek

As of April 2026, DeepSeek's average monthly usage time per person was 109.5 minutes, a 106.9% year-on-year increase.

average monthly usage time per person
109.5 · 截止到2026年4月 · DeepSeek
year-on-year growth
106.9 · 截止到2026年4月 · DeepSeek

Compared metric: m2 — year-on-year growth

QuestMobilesourceaudience usageevidence for deepseek

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

As of April 2026, DeepSeek's average monthly usage time per person was 109.5 minutes, a year-on-year increase of 106.9%.

average monthly usage time per person
109.5 minutes · 截止到2026年4月 · DeepSeek
year-on-year growth in average monthly usage time per person
106.9 % · 截止到2026年4月 · DeepSeek

Compared metric: m2 — year-on-year growth in average monthly usage time per person

QuestMobilesourceaudience usageevidence for deepseek

Deepseek held 0.02% of the AI chatbot market share worldwide in August 2026.

market share
0.02 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for deepseek

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

As of April 2026, DeepSeek's average monthly usage time per person was 109.5 minutes, a year-on-year increase of 106.9%.

average monthly usage time per person
109.5 minutes · 截止到2026年4月 · DeepSeek
year-on-year growth in average monthly usage time per person
106.9 % · 截止到2026年4月 · DeepSeek

Compared metric: m2 — year-on-year growth in average monthly usage time per person

QuestMobilesourceaudience usageevidence for deepseek

As of April 2026, DeepSeek's average monthly usage time per person was 109.5 minutes, a 106.9% year-on-year increase.

average monthly usage time per person
109.5 · 截止到2026年4月 · DeepSeek
year-on-year growth
106.9 · 截止到2026年4月 · DeepSeek

Compared metric: m1 — average monthly usage time per person

QuestMobilesourceaudience usageevidence for deepseek

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

As of April 2026, DeepSeek's average monthly usage time per person was 109.5 minutes, a year-on-year increase of 106.9%.

average monthly usage time per person
109.5 minutes · 截止到2026年4月 · DeepSeek
year-on-year growth in average monthly usage time per person
106.9 % · 截止到2026年4月 · DeepSeek

Compared metric: m2 — year-on-year growth in average monthly usage time per person

QuestMobilesourceaudience usageevidence for deepseek

As of April 2026, DeepSeek's average monthly usage time per person was 109.5 minutes, a 106.9% year-on-year increase.

average monthly usage time per person
109.5 · 截止到2026年4月 · DeepSeek
year-on-year growth
106.9 · 截止到2026年4月 · DeepSeek

Compared metric: m2 — year-on-year growth

QuestMobilesourceaudience usageevidence for deepseek

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Deepseek held 0.02% of the AI chatbot market share worldwide in August 2026.

market share
0.02 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for deepseek

As of April 2026, DeepSeek's average monthly usage time per person was 109.5 minutes, a 106.9% year-on-year increase.

average monthly usage time per person
109.5 · 截止到2026年4月 · DeepSeek
year-on-year growth
106.9 · 截止到2026年4月 · DeepSeek

Compared metric: m1 — average monthly usage time per person

QuestMobilesourceaudience usageevidence for deepseek

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Deepseek held 0.02% of the AI chatbot market share worldwide in August 2026.

market share
0.02 percent · August 2026 · Worldwide

Compared metric: m1 — market share

Statcountersourceaudience usageevidence for deepseek

As of April 2026, DeepSeek's average monthly usage time per person was 109.5 minutes, a 106.9% year-on-year increase.

average monthly usage time per person
109.5 · 截止到2026年4月 · DeepSeek
year-on-year growth
106.9 · 截止到2026年4月 · DeepSeek

Compared metric: m2 — year-on-year growth

QuestMobilesourceaudience usageevidence for deepseek

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

As of April 2026, DeepSeek's average monthly usage time per person was 109.5 minutes, a 106.9% year-on-year increase.

average monthly usage time per person
109.5 · 截止到2026年4月 · DeepSeek
year-on-year growth
106.9 · 截止到2026年4月 · DeepSeek

Compared metric: m1 — average monthly usage time per person

QuestMobilesourceaudience usageevidence for deepseek

As of April 2026, DeepSeek's average monthly usage time per person was 109.5 minutes, a 106.9% year-on-year increase.

average monthly usage time per person
109.5 · 截止到2026年4月 · DeepSeek
year-on-year growth
106.9 · 截止到2026年4月 · DeepSeek

Compared metric: m2 — year-on-year growth

QuestMobilesourceaudience usageevidence for deepseek

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

As of April 2026, Doubao's average monthly usage time per person was 144.6 minutes, a year-on-year increase of 80.6%.

average monthly usage time per person
144.6 minutes · 截止到2026年4月 · 豆包
year-on-year growth in average monthly usage time per person
80.6 % · 截止到2026年4月 · 豆包

Compared metric: m1 — average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

As of April 2026, Doubao's average monthly usage time per person was 144.6 minutes, a year-on-year increase of 80.6%.

average monthly usage time per person
144.6 minutes · 截止到2026年4月 · 豆包
year-on-year growth in average monthly usage time per person
80.6 % · 截止到2026年4月 · 豆包

Compared metric: m2 — year-on-year growth in average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

As of April 2026, Doubao's average monthly usage time per person was 144.6 minutes, a year-on-year increase of 80.6%.

average monthly usage time per person
144.6 minutes · 截止到2026年4月 · 豆包
year-on-year growth in average monthly usage time per person
80.6 % · 截止到2026年4月 · 豆包

Compared metric: m1 — average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

In April 2026, Doubao's average monthly usage time per person was 144.6 minutes, and DeepSeek's was 109.5 minutes, representing year-on-year growth of 80.6% and 106.9% respectively.

Doubao average monthly usage time per person
144.6 minutes
DeepSeek average monthly usage time per person
109.5 minutes
Doubao year-on-year growth in average monthly usage time per person
80.6 %
DeepSeek year-on-year growth in average monthly usage time per person
106.9 %

Compared metric: m2 — DeepSeek average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

As of April 2026, Doubao's average monthly usage time per person was 144.6 minutes, a year-on-year increase of 80.6%.

average monthly usage time per person
144.6 minutes · 截止到2026年4月 · 豆包
year-on-year growth in average monthly usage time per person
80.6 % · 截止到2026年4月 · 豆包

Compared metric: m1 — average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

In April 2026, Doubao's average monthly usage time per person was 144.6 minutes, and DeepSeek's was 109.5 minutes, representing year-on-year growth of 80.6% and 106.9% respectively.

Doubao average monthly usage time per person
144.6 minutes
DeepSeek average monthly usage time per person
109.5 minutes
Doubao year-on-year growth in average monthly usage time per person
80.6 %
DeepSeek year-on-year growth in average monthly usage time per person
106.9 %

Compared metric: m3 — Doubao year-on-year growth in average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

As of April 2026, Doubao's average monthly usage time per person was 144.6 minutes, a year-on-year increase of 80.6%.

average monthly usage time per person
144.6 minutes · 截止到2026年4月 · 豆包
year-on-year growth in average monthly usage time per person
80.6 % · 截止到2026年4月 · 豆包

Compared metric: m1 — average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

In April 2026, Doubao's average monthly usage time per person was 144.6 minutes, and DeepSeek's was 109.5 minutes, representing year-on-year growth of 80.6% and 106.9% respectively.

Doubao average monthly usage time per person
144.6 minutes
DeepSeek average monthly usage time per person
109.5 minutes
Doubao year-on-year growth in average monthly usage time per person
80.6 %
DeepSeek year-on-year growth in average monthly usage time per person
106.9 %

Compared metric: m4 — DeepSeek year-on-year growth in average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

As of April 2026, Doubao's average monthly usage time per person was 144.6 minutes, a year-on-year increase of 80.6%.

average monthly usage time per person
144.6 minutes · 截止到2026年4月 · 豆包
year-on-year growth in average monthly usage time per person
80.6 % · 截止到2026年4月 · 豆包

Compared metric: m1 — average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

As of April 2026, Doubao's average monthly usage time per person was 144.6 minutes, an 80.6% year-on-year increase.

average monthly usage time per person
144.6 · 截止到2026年4月 · 豆包
year-on-year growth
80.6 · 截止到2026年4月 · 豆包

Compared metric: m1 — average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

Agency interpretationThese findings differ in unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • unit

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

As of April 2026, Doubao's average monthly usage time per person was 144.6 minutes, a year-on-year increase of 80.6%.

average monthly usage time per person
144.6 minutes · 截止到2026年4月 · 豆包
year-on-year growth in average monthly usage time per person
80.6 % · 截止到2026年4月 · 豆包

Compared metric: m1 — average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

As of April 2026, Doubao's average monthly usage time per person was 144.6 minutes, an 80.6% year-on-year increase.

average monthly usage time per person
144.6 · 截止到2026年4月 · 豆包
year-on-year growth
80.6 · 截止到2026年4月 · 豆包

Compared metric: m2 — year-on-year growth

QuestMobilesourceaudience usageevidence for doubao

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

As of April 2026, Doubao's average monthly usage time per person was 144.6 minutes, a year-on-year increase of 80.6%.

average monthly usage time per person
144.6 minutes · 截止到2026年4月 · 豆包
year-on-year growth in average monthly usage time per person
80.6 % · 截止到2026年4月 · 豆包

Compared metric: m2 — year-on-year growth in average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

In April 2026, Doubao's average monthly usage time per person was 144.6 minutes, and DeepSeek's was 109.5 minutes, representing year-on-year growth of 80.6% and 106.9% respectively.

Doubao average monthly usage time per person
144.6 minutes
DeepSeek average monthly usage time per person
109.5 minutes
Doubao year-on-year growth in average monthly usage time per person
80.6 %
DeepSeek year-on-year growth in average monthly usage time per person
106.9 %

Compared metric: m1 — Doubao average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

As of April 2026, Doubao's average monthly usage time per person was 144.6 minutes, a year-on-year increase of 80.6%.

average monthly usage time per person
144.6 minutes · 截止到2026年4月 · 豆包
year-on-year growth in average monthly usage time per person
80.6 % · 截止到2026年4月 · 豆包

Compared metric: m2 — year-on-year growth in average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

In April 2026, Doubao's average monthly usage time per person was 144.6 minutes, and DeepSeek's was 109.5 minutes, representing year-on-year growth of 80.6% and 106.9% respectively.

Doubao average monthly usage time per person
144.6 minutes
DeepSeek average monthly usage time per person
109.5 minutes
Doubao year-on-year growth in average monthly usage time per person
80.6 %
DeepSeek year-on-year growth in average monthly usage time per person
106.9 %

Compared metric: m2 — DeepSeek average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

As of April 2026, Doubao's average monthly usage time per person was 144.6 minutes, a year-on-year increase of 80.6%.

average monthly usage time per person
144.6 minutes · 截止到2026年4月 · 豆包
year-on-year growth in average monthly usage time per person
80.6 % · 截止到2026年4月 · 豆包

Compared metric: m2 — year-on-year growth in average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

In April 2026, Doubao's average monthly usage time per person was 144.6 minutes, and DeepSeek's was 109.5 minutes, representing year-on-year growth of 80.6% and 106.9% respectively.

Doubao average monthly usage time per person
144.6 minutes
DeepSeek average monthly usage time per person
109.5 minutes
Doubao year-on-year growth in average monthly usage time per person
80.6 %
DeepSeek year-on-year growth in average monthly usage time per person
106.9 %

Compared metric: m4 — DeepSeek year-on-year growth in average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

As of April 2026, Doubao's average monthly usage time per person was 144.6 minutes, a year-on-year increase of 80.6%.

average monthly usage time per person
144.6 minutes · 截止到2026年4月 · 豆包
year-on-year growth in average monthly usage time per person
80.6 % · 截止到2026年4月 · 豆包

Compared metric: m2 — year-on-year growth in average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

As of April 2026, Doubao's average monthly usage time per person was 144.6 minutes, an 80.6% year-on-year increase.

average monthly usage time per person
144.6 · 截止到2026年4月 · 豆包
year-on-year growth
80.6 · 截止到2026年4月 · 豆包

Compared metric: m1 — average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

As of April 2026, Doubao's average monthly usage time per person was 144.6 minutes, a year-on-year increase of 80.6%.

average monthly usage time per person
144.6 minutes · 截止到2026年4月 · 豆包
year-on-year growth in average monthly usage time per person
80.6 % · 截止到2026年4月 · 豆包

Compared metric: m2 — year-on-year growth in average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

As of April 2026, Doubao's average monthly usage time per person was 144.6 minutes, an 80.6% year-on-year increase.

average monthly usage time per person
144.6 · 截止到2026年4月 · 豆包
year-on-year growth
80.6 · 截止到2026年4月 · 豆包

Compared metric: m2 — year-on-year growth

QuestMobilesourceaudience usageevidence for doubao

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

In April 2026, Doubao's average monthly usage time per person was 144.6 minutes, and DeepSeek's was 109.5 minutes, representing year-on-year growth of 80.6% and 106.9% respectively.

Doubao average monthly usage time per person
144.6 minutes
DeepSeek average monthly usage time per person
109.5 minutes
Doubao year-on-year growth in average monthly usage time per person
80.6 %
DeepSeek year-on-year growth in average monthly usage time per person
106.9 %

Compared metric: m1 — Doubao average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

In April 2026, Doubao's average monthly usage time per person was 144.6 minutes, and DeepSeek's was 109.5 minutes, representing year-on-year growth of 80.6% and 106.9% respectively.

Doubao average monthly usage time per person
144.6 minutes
DeepSeek average monthly usage time per person
109.5 minutes
Doubao year-on-year growth in average monthly usage time per person
80.6 %
DeepSeek year-on-year growth in average monthly usage time per person
106.9 %

Compared metric: m2 — DeepSeek average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

In April 2026, Doubao's average monthly usage time per person was 144.6 minutes, and DeepSeek's was 109.5 minutes, representing year-on-year growth of 80.6% and 106.9% respectively.

Doubao average monthly usage time per person
144.6 minutes
DeepSeek average monthly usage time per person
109.5 minutes
Doubao year-on-year growth in average monthly usage time per person
80.6 %
DeepSeek year-on-year growth in average monthly usage time per person
106.9 %

Compared metric: m1 — Doubao average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

In April 2026, Doubao's average monthly usage time per person was 144.6 minutes, and DeepSeek's was 109.5 minutes, representing year-on-year growth of 80.6% and 106.9% respectively.

Doubao average monthly usage time per person
144.6 minutes
DeepSeek average monthly usage time per person
109.5 minutes
Doubao year-on-year growth in average monthly usage time per person
80.6 %
DeepSeek year-on-year growth in average monthly usage time per person
106.9 %

Compared metric: m3 — Doubao year-on-year growth in average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

In April 2026, Doubao's average monthly usage time per person was 144.6 minutes, and DeepSeek's was 109.5 minutes, representing year-on-year growth of 80.6% and 106.9% respectively.

Doubao average monthly usage time per person
144.6 minutes
DeepSeek average monthly usage time per person
109.5 minutes
Doubao year-on-year growth in average monthly usage time per person
80.6 %
DeepSeek year-on-year growth in average monthly usage time per person
106.9 %

Compared metric: m1 — Doubao average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

In April 2026, Doubao's average monthly usage time per person was 144.6 minutes, and DeepSeek's was 109.5 minutes, representing year-on-year growth of 80.6% and 106.9% respectively.

Doubao average monthly usage time per person
144.6 minutes
DeepSeek average monthly usage time per person
109.5 minutes
Doubao year-on-year growth in average monthly usage time per person
80.6 %
DeepSeek year-on-year growth in average monthly usage time per person
106.9 %

Compared metric: m4 — DeepSeek year-on-year growth in average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

In April 2026, Doubao's average monthly usage time per person was 144.6 minutes, and DeepSeek's was 109.5 minutes, representing year-on-year growth of 80.6% and 106.9% respectively.

Doubao average monthly usage time per person
144.6 minutes
DeepSeek average monthly usage time per person
109.5 minutes
Doubao year-on-year growth in average monthly usage time per person
80.6 %
DeepSeek year-on-year growth in average monthly usage time per person
106.9 %

Compared metric: m1 — Doubao average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

As of April 2026, Doubao's average monthly usage time per person was 144.6 minutes, an 80.6% year-on-year increase.

average monthly usage time per person
144.6 · 截止到2026年4月 · 豆包
year-on-year growth
80.6 · 截止到2026年4月 · 豆包

Compared metric: m1 — average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

In April 2026, Doubao's average monthly usage time per person was 144.6 minutes, and DeepSeek's was 109.5 minutes, representing year-on-year growth of 80.6% and 106.9% respectively.

Doubao average monthly usage time per person
144.6 minutes
DeepSeek average monthly usage time per person
109.5 minutes
Doubao year-on-year growth in average monthly usage time per person
80.6 %
DeepSeek year-on-year growth in average monthly usage time per person
106.9 %

Compared metric: m1 — Doubao average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

As of April 2026, Doubao's average monthly usage time per person was 144.6 minutes, an 80.6% year-on-year increase.

average monthly usage time per person
144.6 · 截止到2026年4月 · 豆包
year-on-year growth
80.6 · 截止到2026年4月 · 豆包

Compared metric: m2 — year-on-year growth

QuestMobilesourceaudience usageevidence for doubao

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

In April 2026, Doubao's average monthly usage time per person was 144.6 minutes, and DeepSeek's was 109.5 minutes, representing year-on-year growth of 80.6% and 106.9% respectively.

Doubao average monthly usage time per person
144.6 minutes
DeepSeek average monthly usage time per person
109.5 minutes
Doubao year-on-year growth in average monthly usage time per person
80.6 %
DeepSeek year-on-year growth in average monthly usage time per person
106.9 %

Compared metric: m2 — DeepSeek average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

In April 2026, Doubao's average monthly usage time per person was 144.6 minutes, and DeepSeek's was 109.5 minutes, representing year-on-year growth of 80.6% and 106.9% respectively.

Doubao average monthly usage time per person
144.6 minutes
DeepSeek average monthly usage time per person
109.5 minutes
Doubao year-on-year growth in average monthly usage time per person
80.6 %
DeepSeek year-on-year growth in average monthly usage time per person
106.9 %

Compared metric: m3 — Doubao year-on-year growth in average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

In April 2026, Doubao's average monthly usage time per person was 144.6 minutes, and DeepSeek's was 109.5 minutes, representing year-on-year growth of 80.6% and 106.9% respectively.

Doubao average monthly usage time per person
144.6 minutes
DeepSeek average monthly usage time per person
109.5 minutes
Doubao year-on-year growth in average monthly usage time per person
80.6 %
DeepSeek year-on-year growth in average monthly usage time per person
106.9 %

Compared metric: m2 — DeepSeek average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

In April 2026, Doubao's average monthly usage time per person was 144.6 minutes, and DeepSeek's was 109.5 minutes, representing year-on-year growth of 80.6% and 106.9% respectively.

Doubao average monthly usage time per person
144.6 minutes
DeepSeek average monthly usage time per person
109.5 minutes
Doubao year-on-year growth in average monthly usage time per person
80.6 %
DeepSeek year-on-year growth in average monthly usage time per person
106.9 %

Compared metric: m4 — DeepSeek year-on-year growth in average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

In April 2026, Doubao's average monthly usage time per person was 144.6 minutes, and DeepSeek's was 109.5 minutes, representing year-on-year growth of 80.6% and 106.9% respectively.

Doubao average monthly usage time per person
144.6 minutes
DeepSeek average monthly usage time per person
109.5 minutes
Doubao year-on-year growth in average monthly usage time per person
80.6 %
DeepSeek year-on-year growth in average monthly usage time per person
106.9 %

Compared metric: m2 — DeepSeek average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

As of April 2026, Doubao's average monthly usage time per person was 144.6 minutes, an 80.6% year-on-year increase.

average monthly usage time per person
144.6 · 截止到2026年4月 · 豆包
year-on-year growth
80.6 · 截止到2026年4月 · 豆包

Compared metric: m1 — average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

In April 2026, Doubao's average monthly usage time per person was 144.6 minutes, and DeepSeek's was 109.5 minutes, representing year-on-year growth of 80.6% and 106.9% respectively.

Doubao average monthly usage time per person
144.6 minutes
DeepSeek average monthly usage time per person
109.5 minutes
Doubao year-on-year growth in average monthly usage time per person
80.6 %
DeepSeek year-on-year growth in average monthly usage time per person
106.9 %

Compared metric: m2 — DeepSeek average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

As of April 2026, Doubao's average monthly usage time per person was 144.6 minutes, an 80.6% year-on-year increase.

average monthly usage time per person
144.6 · 截止到2026年4月 · 豆包
year-on-year growth
80.6 · 截止到2026年4月 · 豆包

Compared metric: m2 — year-on-year growth

QuestMobilesourceaudience usageevidence for doubao

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

In April 2026, Doubao's average monthly usage time per person was 144.6 minutes, and DeepSeek's was 109.5 minutes, representing year-on-year growth of 80.6% and 106.9% respectively.

Doubao average monthly usage time per person
144.6 minutes
DeepSeek average monthly usage time per person
109.5 minutes
Doubao year-on-year growth in average monthly usage time per person
80.6 %
DeepSeek year-on-year growth in average monthly usage time per person
106.9 %

Compared metric: m3 — Doubao year-on-year growth in average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

In April 2026, Doubao's average monthly usage time per person was 144.6 minutes, and DeepSeek's was 109.5 minutes, representing year-on-year growth of 80.6% and 106.9% respectively.

Doubao average monthly usage time per person
144.6 minutes
DeepSeek average monthly usage time per person
109.5 minutes
Doubao year-on-year growth in average monthly usage time per person
80.6 %
DeepSeek year-on-year growth in average monthly usage time per person
106.9 %

Compared metric: m4 — DeepSeek year-on-year growth in average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

In April 2026, Doubao's average monthly usage time per person was 144.6 minutes, and DeepSeek's was 109.5 minutes, representing year-on-year growth of 80.6% and 106.9% respectively.

Doubao average monthly usage time per person
144.6 minutes
DeepSeek average monthly usage time per person
109.5 minutes
Doubao year-on-year growth in average monthly usage time per person
80.6 %
DeepSeek year-on-year growth in average monthly usage time per person
106.9 %

Compared metric: m3 — Doubao year-on-year growth in average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

As of April 2026, Doubao's average monthly usage time per person was 144.6 minutes, an 80.6% year-on-year increase.

average monthly usage time per person
144.6 · 截止到2026年4月 · 豆包
year-on-year growth
80.6 · 截止到2026年4月 · 豆包

Compared metric: m1 — average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

In April 2026, Doubao's average monthly usage time per person was 144.6 minutes, and DeepSeek's was 109.5 minutes, representing year-on-year growth of 80.6% and 106.9% respectively.

Doubao average monthly usage time per person
144.6 minutes
DeepSeek average monthly usage time per person
109.5 minutes
Doubao year-on-year growth in average monthly usage time per person
80.6 %
DeepSeek year-on-year growth in average monthly usage time per person
106.9 %

Compared metric: m3 — Doubao year-on-year growth in average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

As of April 2026, Doubao's average monthly usage time per person was 144.6 minutes, an 80.6% year-on-year increase.

average monthly usage time per person
144.6 · 截止到2026年4月 · 豆包
year-on-year growth
80.6 · 截止到2026年4月 · 豆包

Compared metric: m2 — year-on-year growth

QuestMobilesourceaudience usageevidence for doubao

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

In April 2026, Doubao's average monthly usage time per person was 144.6 minutes, and DeepSeek's was 109.5 minutes, representing year-on-year growth of 80.6% and 106.9% respectively.

Doubao average monthly usage time per person
144.6 minutes
DeepSeek average monthly usage time per person
109.5 minutes
Doubao year-on-year growth in average monthly usage time per person
80.6 %
DeepSeek year-on-year growth in average monthly usage time per person
106.9 %

Compared metric: m4 — DeepSeek year-on-year growth in average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

As of April 2026, Doubao's average monthly usage time per person was 144.6 minutes, an 80.6% year-on-year increase.

average monthly usage time per person
144.6 · 截止到2026年4月 · 豆包
year-on-year growth
80.6 · 截止到2026年4月 · 豆包

Compared metric: m1 — average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

In April 2026, Doubao's average monthly usage time per person was 144.6 minutes, and DeepSeek's was 109.5 minutes, representing year-on-year growth of 80.6% and 106.9% respectively.

Doubao average monthly usage time per person
144.6 minutes
DeepSeek average monthly usage time per person
109.5 minutes
Doubao year-on-year growth in average monthly usage time per person
80.6 %
DeepSeek year-on-year growth in average monthly usage time per person
106.9 %

Compared metric: m4 — DeepSeek year-on-year growth in average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

As of April 2026, Doubao's average monthly usage time per person was 144.6 minutes, an 80.6% year-on-year increase.

average monthly usage time per person
144.6 · 截止到2026年4月 · 豆包
year-on-year growth
80.6 · 截止到2026年4月 · 豆包

Compared metric: m2 — year-on-year growth

QuestMobilesourceaudience usageevidence for doubao

Agency interpretationThese findings differ in metric, unit and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric
  • unit

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

As of April 2026, Doubao's average monthly usage time per person was 144.6 minutes, an 80.6% year-on-year increase.

average monthly usage time per person
144.6 · 截止到2026年4月 · 豆包
year-on-year growth
80.6 · 截止到2026年4月 · 豆包

Compared metric: m1 — average monthly usage time per person

QuestMobilesourceaudience usageevidence for doubao

As of April 2026, Doubao's average monthly usage time per person was 144.6 minutes, an 80.6% year-on-year increase.

average monthly usage time per person
144.6 · 截止到2026年4月 · 豆包
year-on-year growth
80.6 · 截止到2026年4月 · 豆包

Compared metric: m2 — year-on-year growth

QuestMobilesourceaudience usageevidence for doubao

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

NAVER's main page receives an average of 50 million daily visitors, which is the environment for which the AI Tab's next-generation model is optimized.

daily visitors
50000000.0 · average · NAVER’s main page

Compared metric: m1 — daily visitors

NAVERsourceaudience usageevidence for naver-ai-tab

Users who visited NAVER's AI Tab 11 or more times generated twice as many place clicks as users who visited only once.

place clicks
2.0 · users who visited the AI Tab 11 or more times compared to users who visited only once

Compared metric: m1 — place clicks

NAVERsourcereferrals conversionevidence for naver-ai-tab

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

NAVER's main page receives an average of 50 million daily visitors, which is the environment for which the AI Tab's next-generation model is optimized.

daily visitors
50000000.0 · average · NAVER’s main page

Compared metric: m1 — daily visitors

NAVERsourceaudience usageevidence for naver-ai-tab

Users who visited NAVER's AI Tab 11 or more times generated 2.7 times more product clicks than users who visited only once.

product clicks
2.7 · users who visited the AI Tab 11 or more times compared to users who visited only once

Compared metric: m1 — product clicks

NAVERsourcereferrals conversionevidence for naver-ai-tab

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

NAVER's main page receives an average of 50 million daily visitors, which is the environment for which the AI Tab's next-generation model is optimized.

daily visitors
50000000.0 · average · NAVER’s main page

Compared metric: m1 — daily visitors

NAVERsourceaudience usageevidence for naver-ai-tab

During the beta period, click-through rates (CTR) for both product and place cards on NAVER's AI Tab exceeded 20%.

click-through rate
20.0 · During the beta period · for both product and place cards

Compared metric: m1 — click-through rate

NAVERsourcereferrals conversionevidence for naver-ai-tab

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

NAVER's main page receives an average of 50 million daily visitors, which is the environment for which the AI Tab's next-generation model is optimized.

daily visitors
50000000.0 · average · NAVER’s main page

Compared metric: m1 — daily visitors

NAVERsourceaudience usageevidence for naver-ai-tab

NAVER's AI Tab surpassed 4 million cumulative users within two months of its beta launch.

cumulative users
4000000.0 · in about two months · AI Tab, which launched in beta in April for NAVER Plus Membership users

Compared metric: m1 — cumulative users

NAVERsourceaudience usageevidence for naver-ai-tab

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Users who visited NAVER's AI Tab 11 or more times generated twice as many place clicks as users who visited only once.

place clicks
2.0 · users who visited the AI Tab 11 or more times compared to users who visited only once

Compared metric: m1 — place clicks

NAVERsourcereferrals conversionevidence for naver-ai-tab

Users who visited NAVER's AI Tab 11 or more times generated 2.7 times more product clicks than users who visited only once.

product clicks
2.7 · users who visited the AI Tab 11 or more times compared to users who visited only once

Compared metric: m1 — product clicks

NAVERsourcereferrals conversionevidence for naver-ai-tab

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Users who visited NAVER's AI Tab 11 or more times generated twice as many place clicks as users who visited only once.

place clicks
2.0 · users who visited the AI Tab 11 or more times compared to users who visited only once

Compared metric: m1 — place clicks

NAVERsourcereferrals conversionevidence for naver-ai-tab

During the beta period, click-through rates (CTR) for both product and place cards on NAVER's AI Tab exceeded 20%.

click-through rate
20.0 · During the beta period · for both product and place cards

Compared metric: m1 — click-through rate

NAVERsourcereferrals conversionevidence for naver-ai-tab

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Users who visited NAVER's AI Tab 11 or more times generated twice as many place clicks as users who visited only once.

place clicks
2.0 · users who visited the AI Tab 11 or more times compared to users who visited only once

Compared metric: m1 — place clicks

NAVERsourcereferrals conversionevidence for naver-ai-tab

NAVER's AI Tab surpassed 4 million cumulative users within two months of its beta launch.

cumulative users
4000000.0 · in about two months · AI Tab, which launched in beta in April for NAVER Plus Membership users

Compared metric: m1 — cumulative users

NAVERsourceaudience usageevidence for naver-ai-tab

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Users who visited NAVER's AI Tab 11 or more times generated 2.7 times more product clicks than users who visited only once.

product clicks
2.7 · users who visited the AI Tab 11 or more times compared to users who visited only once

Compared metric: m1 — product clicks

NAVERsourcereferrals conversionevidence for naver-ai-tab

During the beta period, click-through rates (CTR) for both product and place cards on NAVER's AI Tab exceeded 20%.

click-through rate
20.0 · During the beta period · for both product and place cards

Compared metric: m1 — click-through rate

NAVERsourcereferrals conversionevidence for naver-ai-tab

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

Users who visited NAVER's AI Tab 11 or more times generated 2.7 times more product clicks than users who visited only once.

product clicks
2.7 · users who visited the AI Tab 11 or more times compared to users who visited only once

Compared metric: m1 — product clicks

NAVERsourcereferrals conversionevidence for naver-ai-tab

NAVER's AI Tab surpassed 4 million cumulative users within two months of its beta launch.

cumulative users
4000000.0 · in about two months · AI Tab, which launched in beta in April for NAVER Plus Membership users

Compared metric: m1 — cumulative users

NAVERsourceaudience usageevidence for naver-ai-tab

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
Contextualisesmethodologically incomparable

Confidence adjustment −0.15

During the beta period, click-through rates (CTR) for both product and place cards on NAVER's AI Tab exceeded 20%.

click-through rate
20.0 · During the beta period · for both product and place cards

Compared metric: m1 — click-through rate

NAVERsourcereferrals conversionevidence for naver-ai-tab

NAVER's AI Tab surpassed 4 million cumulative users within two months of its beta launch.

cumulative users
4000000.0 · in about two months · AI Tab, which launched in beta in April for NAVER Plus Membership users

Compared metric: m1 — cumulative users

NAVERsourceaudience usageevidence for naver-ai-tab

Agency interpretationThese findings differ in metric and cannot be reconciled as estimates of the same quantity. Preserve both findings in their stated contexts; neither establishes a universal change or absence of change.
Comparison context and methodology

Differences that explain the relationship

  • metric

Unknown context dimensions

  • denominator
  • geography
  • sampling frame

Unresolved

247

Missing methodology — neither agreement nor contradiction can be concluded.

ContextualisesunresolvedContested

Confidence adjustment −0.15

Google's AI Overviews have significantly increased their inclusion of external links, rising from nearly 0% to over 25% of AI Overviews.

percentage of AI Overviews containing external links
0.0 % · AI Overviews
percentage of AI Overviews containing external links
25.0 % · AI Overviews

Compared metric: m1 — percentage of AI Overviews containing external links

Search Engine Roundtablesourcecitations sourcesevidence for google-ai-overviews

Google's AI Overviews have significantly increased their inclusion of external links, rising from nearly 0% to over 25% of AI Overviews.

percentage of AI Overviews containing external links
0.0 % · AI Overviews
percentage of AI Overviews containing external links
25.0 % · AI Overviews

Compared metric: m2 — percentage of AI Overviews containing external links

Search Engine Roundtablesourcecitations sourcesevidence for google-ai-overviews

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (0.0 vs 25.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

The click-through rate for a search result at position 1 is reduced from 27% to 11% when an AI Overview is present.

click-through rate
27.0 · position 1 search result without AI Overview
click-through rate
11.0 · position 1 search result with AI Overview

Compared metric: m1 — click-through rate

SISTRIXsourceoptimisation implicationevidence for google-ai-overviews

The click-through rate for a search result at position 1 is reduced from 27% to 11% when an AI Overview is present.

click-through rate
27.0 · position 1 search result without AI Overview
click-through rate
11.0 · position 1 search result with AI Overview

Compared metric: m2 — click-through rate

SISTRIXsourceoptimisation implicationevidence for google-ai-overviews

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (27.0 vs 11.0); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Reddit's citation share also fell in Google's AI Overviews and AI Mode over the same period, by 11% and 31% respectively, which is shallower than the ChatGPT drop.

fall in citation share
11.0 · over the same period · in Google's AI Overviews
fall in citation share
31.0 · over the same period · in Google's AI Mode

Compared metric: m1 — fall in citation share

Semrushsourcecitations sourcesevidence for google-ai-overviews

Reddit's citation share also fell in Google's AI Overviews and AI Mode over the same period, by 11% and 31% respectively, which is shallower than the ChatGPT drop.

fall in citation share
11.0 · over the same period · in Google's AI Overviews
fall in citation share
31.0 · over the same period · in Google's AI Mode

Compared metric: m2 — fall in citation share

Semrushsourcecitations sourcesevidence for google-ai-overviews

Agency interpretationInsufficient context to compare: denominator, geography, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (11.0 vs 31.0); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline element (image embeddings) rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0
share of responses with inline elements
35.0

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (15.0 vs 35.0); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image element rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline elements
15.0 % · before May 7-11, 2026 · ChatGPT responses
share of responses with inline elements
35.0 % · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline elements

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (15.0 vs 35.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-19.0 vs -22.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-19.0 vs -22.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-19.0 vs -73.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-19.0 vs -66.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-19.0 vs -34.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-19.0 vs -60.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-19.0 vs -60.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-19.0 vs -37.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-19.0 vs -52.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-19.0 vs -17.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-19.0 vs -14.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-19.0 vs -47.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-19.0 vs -18.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Yahoo lost 19% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-19.0 % · Yahoo in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-19.0 vs 59.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-22.0 vs -73.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-22.0 vs -66.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-22.0 vs -34.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-22.0 vs -60.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-22.0 vs -60.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-22.0 vs -37.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-22.0 vs -52.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-22.0 vs -17.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-22.0 vs -14.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-22.0 vs -47.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-22.0 vs -18.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

LinkedIn lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · LinkedIn in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-22.0 vs 59.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-22.0 vs -73.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-22.0 vs -66.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-22.0 vs -34.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-22.0 vs -60.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-22.0 vs -60.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-22.0 vs -37.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-22.0 vs -52.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-22.0 vs -17.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-22.0 vs -14.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-22.0 vs -47.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-22.0 vs -18.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Google.com lost 22% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-22.0 % · Google.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-22.0 vs 59.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-73.0 vs -66.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-73.0 vs -34.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-73.0 vs -60.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-73.0 vs -60.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-73.0 vs -37.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-73.0 vs -52.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-73.0 vs -17.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-73.0 vs -14.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-73.0 vs -47.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-73.0 vs -18.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Classic.com lost 73% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-73.0 % · Classic.com in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-73.0 vs 59.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-66.0 vs -34.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-66.0 vs -60.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-66.0 vs -60.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-66.0 vs -37.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-66.0 vs -52.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-66.0 vs -17.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-66.0 vs -14.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-66.0 vs -47.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-66.0 vs -18.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Mobile.de lost 66% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-66.0 % · Mobile.de in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-66.0 vs 59.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-34.0 vs -60.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-34.0 vs -60.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-34.0 vs -37.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-34.0 vs -52.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-34.0 vs -17.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-34.0 vs -14.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-34.0 vs -47.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-34.0 vs -18.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

ZipRecruiter lost 34% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-34.0 % · ZipRecruiter in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-34.0 vs 59.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-60.0 vs -37.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-60.0 vs -52.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-60.0 vs -17.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-60.0 vs -14.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-60.0 vs -47.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-60.0 vs -18.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Rome2rio lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Rome2rio in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-60.0 vs 59.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-60.0 vs -37.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-60.0 vs -52.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-60.0 vs -17.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-60.0 vs -14.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-60.0 vs -47.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-60.0 vs -18.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Expedia lost 60% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-60.0 % · Expedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-60.0 vs 59.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-37.0 vs -52.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-37.0 vs -17.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-37.0 vs -14.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-37.0 vs -47.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-37.0 vs -18.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-37.0 vs 59.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-52.0 vs -17.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-52.0 vs -14.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-52.0 vs -47.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-52.0 vs -18.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Glassdoor lost between 37% and 52% of its citations across all country versions in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-37.0 % · Glassdoor (all country versions) in German-language ChatGPT responses
relative change in citations
-52.0 % · Glassdoor (all country versions) in German-language ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-52.0 vs 59.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-17.0 vs -14.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-17.0 vs -47.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-17.0 vs -18.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Facebook lost 17% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-17.0 % · Facebook in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-17.0 vs 59.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-14.0 vs -47.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-14.0 vs -18.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Wikipedia lost 14% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-14.0 % · Wikipedia in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-14.0 vs 59.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-47.0 vs -18.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Indeed lost 47% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-47.0 % · Indeed in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-47.0 vs 59.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

YouTube lost 18% of its citations in German-language ChatGPT responses after the GPT-5.5 rollout.

relative change in citations
-18.0 % · YouTube in German-language ChatGPT responses

Compared metric: m1 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Reddit gained 7,007 citations per 10,000 ChatGPT responses, a 59% increase, after the GPT-5.5 rollout.

citations
7007.0 citations per 10,000 responses · Reddit in ChatGPT responses
relative change in citations
59.0 % · Reddit in ChatGPT responses

Compared metric: m2 — relative change in citations

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (-18.0 vs 59.0 %); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m1 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources per ChatGPT response dropped from 30.9 to 28.4 after the GPT-5.5 rollout.

average sources per response
30.9 sources · ChatGPT responses before GPT-5.5 rollout
average sources per response
28.4 sources · ChatGPT responses after GPT-5.5 rollout

Compared metric: m2 — average sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (30.9 vs 28.4 sources); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (66.0 vs 51.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (66.0 vs 60.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (66.0 vs 22.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (66.0 vs 19.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (66.0 vs 73.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (66.0 vs 22.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (66.0 vs 34.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (66.0 vs 60.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (66.0 vs 46.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (66.0 vs 52.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (66.0 vs 37.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (66.0 vs 17.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (66.0 vs 53.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (66.0 vs 14.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (66.0 vs 47.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Mobile.de citations in German-language ChatGPT responses decreased by 66% after the GPT-5.5 rollout.

citation decrease
66.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (66.0 vs 18.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (51.0 vs 60.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (51.0 vs 22.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (51.0 vs 19.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (51.0 vs 73.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (51.0 vs 22.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (51.0 vs 34.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (51.0 vs 60.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (51.0 vs 46.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (51.0 vs 52.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (51.0 vs 37.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (51.0 vs 17.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (51.0 vs 53.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (51.0 vs 14.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (51.0 vs 47.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Tripadvisor.co.uk citations in German-language ChatGPT responses decreased by 51% after the GPT-5.5 rollout.

citation decrease
51.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (51.0 vs 18.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (60.0 vs 22.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (60.0 vs 19.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (60.0 vs 73.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (60.0 vs 22.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (60.0 vs 34.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (60.0 vs 46.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (60.0 vs 52.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (60.0 vs 37.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (60.0 vs 17.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (60.0 vs 53.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (60.0 vs 14.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (60.0 vs 47.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Expedia.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (60.0 vs 18.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (22.0 vs 19.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (22.0 vs 73.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (22.0 vs 34.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (22.0 vs 60.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (22.0 vs 46.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (22.0 vs 52.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (22.0 vs 37.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (22.0 vs 17.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (22.0 vs 53.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (22.0 vs 14.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (22.0 vs 47.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Linkedin.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (22.0 vs 18.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (19.0 vs 73.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (19.0 vs 22.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (19.0 vs 34.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (19.0 vs 60.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (19.0 vs 46.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (19.0 vs 52.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (19.0 vs 37.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (19.0 vs 17.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (19.0 vs 53.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (19.0 vs 14.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (19.0 vs 47.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Yahoo.com citations in German-language ChatGPT responses decreased by 19% after the GPT-5.5 rollout.

citation decrease
19.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (19.0 vs 18.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (73.0 vs 22.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (73.0 vs 34.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (73.0 vs 60.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (73.0 vs 46.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (73.0 vs 52.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (73.0 vs 37.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (73.0 vs 17.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (73.0 vs 53.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (73.0 vs 14.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (73.0 vs 47.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Classic.com citations in German-language ChatGPT responses decreased by 73% after the GPT-5.5 rollout.

citation decrease
73.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (73.0 vs 18.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (22.0 vs 34.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (22.0 vs 60.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (22.0 vs 46.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (22.0 vs 52.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (22.0 vs 37.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (22.0 vs 17.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (22.0 vs 53.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (22.0 vs 14.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (22.0 vs 47.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Google.com citations in German-language ChatGPT responses decreased by 22% after the GPT-5.5 rollout.

citation decrease
22.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (22.0 vs 18.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (34.0 vs 60.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (34.0 vs 46.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (34.0 vs 52.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (34.0 vs 37.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (34.0 vs 17.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (34.0 vs 53.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (34.0 vs 14.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (34.0 vs 47.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Ziprecruiter.com citations in German-language ChatGPT responses decreased by 34% after the GPT-5.5 rollout.

citation decrease
34.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (34.0 vs 18.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (60.0 vs 46.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (60.0 vs 52.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (60.0 vs 37.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (60.0 vs 17.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (60.0 vs 53.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (60.0 vs 14.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (60.0 vs 47.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Rome2rio.com citations in German-language ChatGPT responses decreased by 60% after the GPT-5.5 rollout.

citation decrease
60.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (60.0 vs 18.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (46.0 vs 52.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (46.0 vs 37.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (46.0 vs 17.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (46.0 vs 53.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (46.0 vs 14.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (46.0 vs 47.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Kununu.com citations in German-language ChatGPT responses decreased by 46% after the GPT-5.5 rollout.

citation decrease
46.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (46.0 vs 18.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (52.0 vs 37.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (52.0 vs 17.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (52.0 vs 53.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (52.0 vs 14.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (52.0 vs 47.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Glassdoor.co.uk citations in German-language ChatGPT responses decreased by 52% after the GPT-5.5 rollout.

citation decrease
52.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (52.0 vs 18.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (37.0 vs 17.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (37.0 vs 53.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (37.0 vs 14.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (37.0 vs 47.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Glassdoor.com citations in German-language ChatGPT responses decreased by 37% after the GPT-5.5 rollout.

citation decrease
37.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (37.0 vs 18.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (17.0 vs 53.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (17.0 vs 14.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (17.0 vs 47.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Facebook citations in German-language ChatGPT responses decreased by 17% after the GPT-5.5 rollout.

citation decrease
17.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (17.0 vs 18.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (53.0 vs 14.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (53.0 vs 47.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Tripadvisor.com citations in German-language ChatGPT responses decreased by 53% after the GPT-5.5 rollout.

citation decrease
53.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (53.0 vs 18.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (14.0 vs 47.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Wikipedia citations in German-language ChatGPT responses decreased by 14% after the GPT-5.5 rollout.

citation decrease
14.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (14.0 vs 18.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

Indeed.com citations in German-language ChatGPT responses decreased by 47% after the GPT-5.5 rollout.

citation decrease
47.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

YouTube citations in German-language ChatGPT responses decreased by 18% after the GPT-5.5 rollout.

citation decrease
18.0 percent · May 18-21 to May 26-29, 2026 · German-language ChatGPT responses

Compared metric: m1 — citation decrease

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (47.0 vs 18.0 percent); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m1 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average sources per response
30.9 number · before May 22/23, 2026
average sources per response
28.4 number · after May 22/23, 2026

Compared metric: m2 — average sources per response

SISTRIXsourceretrieval indexevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (30.9 vs 28.4 number); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
  • period start
  • period end
ContextualisesunresolvedContested

Confidence adjustment −0.15

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m1 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

The share of ChatGPT responses with at least one inline image embedding from images.openai.com rose from approximately 15% to 35% between May 7-11, 2026.

share of responses with inline image embeddings
15.0 · before May 7-11, 2026 · ChatGPT responses
share of responses with inline image embeddings
35.0 · after May 7-11, 2026 · ChatGPT responses

Compared metric: m2 — share of responses with inline image embeddings

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (15.0 vs 35.0); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m1 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

The average number of sources cited per ChatGPT response dropped from 30.9 to 28.4 during the GPT-5.5 rollout on May 22/23, 2026.

average number of sources per response
30.9 · before GPT-5.5 rollout · ChatGPT responses
average number of sources per response
28.4 · after GPT-5.5 rollout · ChatGPT responses

Compared metric: m2 — average number of sources per response

SISTRIXsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, mode, sampling_frame. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (30.9 vs 28.4); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • mode
  • sampling frame
ContextualisesunresolvedContested

Confidence adjustment −0.15

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m1 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

A study of more than 230,000 prompts tracked a similar collapse in ChatGPT's Reddit citations a year ago, falling from about 60% of responses in early August 2025 to roughly 10% by mid-September.

Reddit citations
60.0 · early August 2025 · of responses
Reddit citations
10.0 · mid-September · of responses

Compared metric: m2 — Reddit citations

Semrushsourcecitations sourcesevidence for chatgpt

Agency interpretationInsufficient context to compare: denominator, geography, sampling_frame, period_start, period_end. Missing methodology is not evidence of agreement or contradiction. Note: known values disagree (60.0 vs 10.0); may indicate a genuine conflict once contexts are completed.
Comparison context and methodology

Differences that explain the relationship

No dimension difference recorded.

Unknown context dimensions

  • denominator
  • geography
  • sampling frame
  • period start
  • period end