AI Discovery Intelligence
How AI discovery works — and what it means for you
A source-backed read on the consumer AI platforms that now answer questions, how each finds, retrieves and cites information, how strong the evidence is, and what a marketer should do about it.
Start here
Four questions, answered
What AI discovery platforms exist?
7 major consumer AI discovery surfaces, with vendor, geography, discovery modes and evidenced reach.
Registry facts are labelled; reach is evidenced.
See the landscape →How does their search and discovery work?
The mechanics source is unavailable right now, so the comparison cannot be shown.
Every dimension is evidenced or explicitly unknown.
Compare how discovery works →How do we know — and can we trust it?
The evidence source is unavailable right now, so findings cannot be shown.
Vendor documentation, independent research and direct observation stay distinct.
Inspect evidence & trust →Why does this matter for marketers?
The implications source is unavailable right now, so actions cannot be shown.
No generic advice: every action cites its supporting claims.
See marketing implications →
Returning users
Latest changes and the weekly brief
What changed recently and what the standing position currently says. Useful intelligence for returning readers, not the front door.
Latest material changes
What changed
The newest persisted change events against the surface registry, newest first.
Showing the newest 5 of 100 changes. All changes in Explore surfaces →
Weekly executive brief
What matters this week
Window 2026-09-20 → 2026-09-27 · generated 2026-09-27
A survey of 2,338 US consumers found that 4 in 10 AI users dislike chatbot ads.
Why it matters & what to do
- Why it matters
- Where discovery carries shopping or ad placements, it is also a commercial pipeline rather than a citation only.
- Agency action
- Align feed, product-data and paid placement strategy to the surface's commerce behaviour.
AI chatbots talked 57.5% of AI users out of buying.
Why it matters & what to do
- Why it matters
- Referral and measurement behaviour determines whether AI discovery can be attributed and defended commercially.
- Agency action
- Update measurement and attribution to capture the referral path where it is observable.
A survey of 2,338 US consumers found that 65% of AI users have replaced some product-related Google searches with chatbots.
Why it matters & what to do
- Why it matters
- A shift in where audiences actually spend attention changes which surfaces are commercially worth prioritising.
- Agency action
- Re-weight platform and channel investment where the measured audience has moved.
Current POV
What we currently believe
Source pov/state.yaml · version 1 · 1 change recorded
- pov-retrieval-systemsconfidence: Medium
How retrieval/search/citation systems differ
Each surface reaches the live web through a different retrieval and citation stack, so the same brand asset is discovered unevenly. Treat retrieval/index and crawler/index documentation as first-class evidence: it changes what "being found" means on that surface.
Channel/surface prioritisation by geography/category
Prioritise by measured audience and referral behaviour, not by press volume. A surface matters when it moves a category's discovery, and that varies by geography; a regional surface can be commercially material before it is globally large.
Commercial surfaces: shopping, transaction and ads
Where AI surfaces carry shopping, transaction or ad placements, discovery is also a commercial pipeline. Optimise the product and feed data those surfaces consume, and measure the commerce path, not just the citation.
Measurement principles and known blind spots
Measure discovery with evidence-derived confidence and explicit unknowns; a vendor's self-reported share is a study with a denominator and a window, not a fact. Known blind spots (unnamed samples, unpublished denominators, different geographies) are recorded, not smoothed over.
Recent POV change
pov-retrieval-systems2026-09-19 18:33 UTC material crawler_index_policy change: significance 3.63 >= 3.50, confidence medium (first-seen quantified claim)
Before → after
Before
Each surface reaches the live web through a different retrieval and citation stack, so the same brand asset is discovered unevenly. Treat retrieval/index and crawler/index documentation as first-class evidence: it changes what "being found" means on that surface.
After
Each surface reaches the live web through a different retrieval and citation stack, so the same brand asset is discovered unevenly. Treat retrieval/index and crawler/index documentation as first-class evidence: it changes what "being found" means on that surface. - It can take approximately 24 hours for OpenAI's systems to adjust for search results after a site's robots.txt update. (evidence 0b187a5b6f942c1d2a3fcb12285238e5; 2026-09-19; medium)
Evidence: 0b187a5b6f94