Brands named per AI answer · median
5
Mean 6.4, across 1,382 answers that named at least one brand.
AI Visibility Report · September 2026
Two engines answering the same buying question cite completely different sources 46% of the time.
Not partly different. Zero overlap. Of the buying questions answered by two or more AI engines this month, 356 could be compared engine-to-engine; 46% shared no cited source at all (95% CI 40-51%). Each engine reads its own web. Check your brand on every engine your buyers use.
Measured on
- 46% of buying questions answered by two or more AI engines shared zero cited sources between engines.
- The median AI answer names 5 brands. No ads, no page two.
- 81% of 7,461 AI citations point at third-party pages the recommended brand does not control.
- Wikipedia earned under 0.4% of citations to buyer questions. Reddit and YouTube dominate.
- When engines cite a brand’s own site, they name the brand 100% of the time (0% ghost citations, n=106).
- 0 sponsored units found inside 35 ad-measurable AI answers. The trend line starts at zero.
Free to cite with a link to this page · every figure detailed in the sections below
The shortlist is five brands long
A first page of Google has 10 slots and ads. An AI answer has a fixed shortlist, no ads, and no page two.
Slots on a Google first page
10
Plus ads. The old game had more seats; the new one has five.
What to do with this
Track whether you are IN the answer, not where you rank in it. Brand six is invisible.
Most-cited sources this month
Every source AI engines attached to their answers, ranked. Engine-owned properties (google.com results pages, Play Store) excluded. Logos = which engines cited the domain.
reddit.com140youtube.com100TRtechradar.com62
62inlinkedin.com47
47Wen.wikipedia.org28
28Mmedium.com27
27Cclutch.co24
24Fforbes.com22
22PRprnewswire.com20
20Ddev.to19
197,461 citations collected from 299 scans of real buyer questions · counts are per-domain citation totals in the window
Two findings hide in the logos. Reddit tops the list while only Google AI Overviews and Perplexity cite it; ChatGPT and Claude barely touch it. And Wikipedia earned 28 citations of 7,461. Under 0.4%. A Wikipedia page is not a GEO strategy.
What to do with this
Earn presence where engines already look: category roundups, review platforms and the Reddit threads your buyers read, before investing anywhere exotic.
Who owns the citations
Of every page engines cited, how many does the recommended brand actually control?
81% earned19% owned
Earned: third-party pages (reviews, forums, media)Owned: pages a named brand controls
n=7,461 citations, 95% CI on earned 80-82% · “owned” = the cited page sits on a site belonging to a brand named in that answer · counted conservatively, so owned is if anything an undercount
What to do with this
Your blog alone cannot win this channel. What other people publish about you decides 4 of every 5 citations.
Four engines, four different webs
How much each engine cites, and where. The volumes alone explain most visibility differences.
PerplexityThe citation firehose: more footnotes than the other three combined. The main reason Reddit tops the leaderboard.4,018ChatGPTBroad media diet: tech press, wire services, publisher sites. The only engine citing PR Newswire this month.2,328ClaudeFewest answers with citations, longest tail: developer blogs and documentation over mass media.1,544AI OverviewsThe most selective: an order of magnitude fewer citations, concentrated on YouTube and community sources.248Citations per engine in the window, engine-owned domains excluded · YouTube and Wikipedia are the only sources all four engines agreed on
What to do with this
One playbook per engine. Reddit presence moves Perplexity and AI Overviews; press moves ChatGPT; technical depth moves Claude.
Who engines trust, by industry
The most-cited external source for buyer questions in each industry (5-site minimum per industry; industry = onboarding survey where present, else inferred and labeled as such).
SaaSmedium.com, techradar.com and dev.to lead; reddit.com close behind
E-commercereddit.com and youtube.com lead; marketplaces (etsy.com) follow
Agencies & consultingclutch.co is the trust anchor, then linkedin.com
Healthcaredirectory authority: betterhelp.com, psychologytoday.com
Travel & hospitalityreddit.com plus the app stores
Media & publishingnytimes.com and en.wikipedia.org, the one industry where legacy authority holds
88 industry-classified sites with completed scans this window; industries below the 5-site floor suppressed · classification backfill still running, cuts deepen next month
What to do with this
Your industry has a trust anchor. Find yours in the row above and make sure you exist there before anywhere else.
The ghost-citation myth
Industry studies claim up to 62% of citations never name the cited brand. We measured it.
0%
Of 106 answers that cited a brand’s own site, zero failed to name the brand in the answer text (95% CI 0-3%). When an engine trusts your page enough to cite it, it credits you.
What to do with this
The fix is upstream, not in the answer: the battle is getting cited at all, not getting credited once cited. Spend accordingly.
Ads inside AI answers: the zero baseline
Sponsored units inside the answer itself, measured on the lanes that expose them.
Sponsored units found · September
0
Across 35 ad-measurable answers (95% CI 0-10%). Engines with no ads lane are excluded, never counted as ad-free.
Why publish a zero
the trend line starts here
When ads arrive inside answers, this series will show exactly when and how fast.
What to do with this
The free-answer era is the buying window. Organic presence earned now becomes the baseline advertisers will have to buy their way past.
Landing in October
Instrumentation that is live and accruing but below our evidence floors this month. We publish when the data clears the bar, not before.
The searches engines run before they answer (query fan-out)capturing since Sep 15
Which trust platforms engines consult but never citecapturing since Sep 15
AI crawler traffic on real sites, by botlogging since Sep 14
Which AI assistants send human visitors, from GA4refreshing now
How to read this
Every number carries its denominator. Real buyer questions asked live to four engines on a schedule across 12 locales; 7,461 citations from 299 scans this window. The panel skews toward commercial intent and the industries our users bring us. This is not a random sample of the web, and we do not claim it is.
Anything below our evidence floor is suppressed rather than published; the October section above is exactly that honesty made visible. Methodology in full at seomatic.ai/research.
Use this data
Every number here is free to cite, quote or republish with a link to this page. Charts and the summary card may be reused with attribution. No permission needed.
AI Visibility Report, September 2026. SEOmatic Research. https://seomatic.ai/research/ai-visibility
<a href="https://seomatic.ai/research/ai-visibility"><img src="https://seomatic.ai/research/og/og-ai-visibility.png" alt="AI Visibility Report, September 2026. SEOmatic Research" width="600"></a>
Where do AI engines send your buyers?
Run this same report on your own domain and see which answers name you, which name your competitors, and why.
Run this report on my domain Free · no credit card · results in about a minute