Ask an AI which brand to pick and it doesn't take that brand's word for it. It leans on whatever it already trusts across the web. We wanted to see exactly what that is. So we measured it: 129 businesses, 5,083 real buyer questions, three engines, about 51,000 sources cited.
AI describes you through other people's pages
We checked every citation for one thing: did it actually point back to the brand being asked about? Pooled together, a brand's own pages were about 12% of its citations. But a handful of heavy brands can carry a number like that, so we redid it one business at a time, one vote each. The median came out at 6.7%.
Then we tried to break it. Drop the three most-cited brands: 11.7%. Drop the top ten: 13%. Barely a wobble. This isn't a couple of big names skewing the average. AI builds its picture of you mostly from pages you don't own, and that holds however we cut it.
Put plainly: you can have a flawless website and still be almost missing from the evidence AI uses to recommend anyone.
AI visibility isn't low, it's wildly uneven
It's tempting to write "the average brand gets mentioned about a third of the time" and move on. That average lies. Measured per business, the middle sat near 36%. But the range was enormous: the bottom quarter scraped around 10%, the top quarter cleared 68%. And one in five businesses never showed up at all, on any engine, on any question.
The variance is the finding
This is not a story about everyone being a little behind. It is a large gap between the brands AI reaches for and the brands it never names, with most sitting closer to the wrong end.
The types of source that get cited
The specific sites change by country. The kind of source doesn't. Past a brand's own pages, the third-party sources you can actually earn fall into a familiar set:
| Source type | What it looks like |
|---|---|
| Editorial | Articles and "best of" roundups that cover your category |
| Review platforms | Sites where customers and users rate you |
| Business directories | Local and category listings and aggregators |
| Social & community | Social profiles and community threads, including Reddit |
JustDial in India, Yelp in the US, some regional directory in Europe. Different site, same pattern: AI trusts aggregators, reviews and editorial coverage to vouch for you. You earn those. You can't publish them yourself.
The three engines don't cite the same way
Normalised per engine, they lean differently:
| Engine | Leans most on |
|---|---|
| Gemini | Directories and aggregators |
| ChatGPT | Official and brand pages, plus reference sources like Wikipedia |
| Perplexity | Social and community sources |
So one "AEO strategy" won't cover all three. Getting cited on Gemini looks nothing like getting cited on ChatGPT. We got into the mechanics in How AI Engines Actually Choose Which Brands to Recommend.
When AI does mention you, it's almost always positive
Of the mentions we scored for sentiment, about 90% were positive and under 4% negative. For most businesses this isn't a reputation problem. AI isn't running you down. It just isn't bringing you up. That's the easier problem to fix, and it points the work off your own site, toward earning the citations above.
What to actually do with this
1. Measure per engine
A win on one engine is not a win on the others. Check ChatGPT, Gemini and Perplexity separately.
2. Earn the right types of third-party sources
Directories, review platforms and editorial coverage for your market. That is where the large majority of your citations will come from. For local brands it starts with an accurate, active Google Business Profile and clean listings.
3. Keep your own pages strong too
Your site is still your most-cited single source, just not the majority. Fix it, then go earn the rest.
See what AI cites about you
BizLoc8 runs your buyer questions across ChatGPT, Gemini and Perplexity and shows exactly which sources each engine pulls to describe you, and which ones you're missing. It's the same audit that produced this study.
How we did this, and what we don't claim
Figures come from first-party BizLoc8 audit data, January to July 2026: 129 businesses, 5,083 prompt results and roughly 51,000 citations across ChatGPT, Gemini and Perplexity. We normalised per business (one business, one vote) so a few heavy brands do not dominate, and we report source types rather than specific domains because the specifics are market-dependent. This audit base skews to India and Southeast Asia and to travel, retail and local-services verticals, so we do not claim a global average. We also avoid stating a fixed "number of sources per answer" because that reflects our capture limit, not the engines. Treat this as directional and structural, not a census.Frequently Asked Questions
What sources does AI cite when it describes a brand?
Mostly third-party ones. In our study of 51,000 citations, the typical business saw only about 7% of its AI citations come from its own website. The rest came from editorial articles, review platforms, business directories, and social or community sites. AI builds its picture of you largely from sources you do not own.
Does my own website still matter for AI visibility?
Yes, it is still your single most-cited source, but it is not the majority. For the typical brand it was a small share of total citations, so a great website alone is not enough. You also have to earn presence on the third-party sources AI trusts.
Do ChatGPT, Gemini and Perplexity cite different sources?
Yes. Normalised per engine, Gemini leans most on directories and aggregators, ChatGPT leans on official and reference sources including Wikipedia, and Perplexity leans more on social and community sources. A single strategy will not cover all three.
Is AI saying negative things about brands?
Rarely. When a brand was mentioned, about 90% of those mentions were positive and under 4% negative. For most businesses the problem is absence, not reputation. AI is not running you down, it simply is not bringing you up.
Related Articles
