AI Brand Mentions vs Citations: Why ChatGPT Can Recommend You and Still Not Cite Your Site
When people ask me about AI brand mentions vs citations, my answer is that they're two different numbers and you need both of them, reported separately, engine by engine. Semrush's new AI Visibility Index found that fewer than half of the brands ChatGPT, Gemini and Google's AI products mention most often are also among the domains they cite most often, and that was true in every month they measured, in both the UK and the US (Semrush, October 13, 2026).
So if your AI visibility report is one number, it's probably telling you about one of those two things and not the other. A lot of the tools out there lean toward citations, because a cited URL is easy to count, and that's fine as far as it goes. It just isn't the same as whether the engine told a buyer to go with you.
The short version
A mention is your brand named in the answer. A citation is a page the engine linked as a source. Semrush found those two lists overlap less than half the time.
The gap is wider on some engines than others. On ChatGPT in the UK, only about 29 to 32 of the top 100 mentioned brands were also in the top 100 cited domains.
Mentions are closer to the verdict a buyer reads. Citations tell you where the engine went for evidence, and they're where any click comes from.
Track both, per engine, and decide which gap you have before you decide what to fix.
What Semrush measured
The study is the second edition of Semrush's AI Visibility Index, released at their Spotlight event in London. They ran fixed monthly sets of 2.5 million UK prompts and 25 million US prompts through ChatGPT, Gemini and Google AI Mode from January through July 2026, and added a separate keyword set for Google AI Overviews (Semrush press release). The full 57-page report is on their AI Visibility Index site.
For each category they compared two lists, the top 100 brands the engines named in their answers and the top 100 domains the engines cited as sources. In the UK the average overlap ran from 36.7 to 45.3 domains out of 100, depending on the month. In the US it was 41.4 to 47.4 (PPC Land summary of the report).
ChatGPT was near the bottom. In the UK, 28.7 to 32.3 of its top 100 mentioned brands showed up in its top 100 cited domains, and in the US it was 32.8 to 35.8 (Relevant Audience). Google AI Overviews had the most, somewhere between about 45 and 64 depending on the month and the market. That makes some sense to me, since AI Overviews sits on top of Google's index and leans on the same pages that rank, though that's my read and not something the report claims.
Named vs. cited
Out of each engine's top 100 most-mentioned brands, how many were also among its top 100 most-cited domains? Bars show the range across seven months (Jan to Jul 2026). The dashed line is 50.
Source: Semrush AI Visibility Index 2026, H2 edition (Oct 13, 2026), via Semrush, PPC Land and Relevant Audience.
The Tesco and Asda example
The example everybody's going to quote from this report is the UK grocery one, and I think it's a good one because the two brands are so evenly matched on the mention side.
In July, ChatGPT ranked Asda and Tesco almost side by side on mentions. Asda was the 8th most-mentioned brand in ChatGPT's UK answers and Tesco was 9th (PPC Land). On citations it wasn't close. Tesco was ChatGPT's 25th most-cited UK domain and Asda was 606th. On Gemini the gap shrank a lot, with Tesco at 29th and Asda at 60th for citations (Semrush).
So say you're on the Asda marketing team and your tool only tracks citations. You'd come away thinking ChatGPT barely knows you, when it names you about as often as it names Tesco and rarely uses your own site as a source when it does. A mentions-only tool would get you the opposite mistake, where everything looks fine and you never see that the engine is building its picture of your brand mostly out of other people's pages.
Same mentions, different citations
Tesco and Asda in AI answers for UK prompts, July 2026. Lower rank is better.
Tesco
Asda
Source: Semrush AI Visibility Index 2026, H2 edition. Citation ranks from Semrush; mention ranks as reported by PPC Land.
Why the two numbers drift apart
I don't think anyone fully knows why, and the report is careful to say it measures recurrence and doesn't prove cause. Still, a few things in the data point the same direction.
The engines behave differently. ChatGPT cites the most sources per answer and Gemini cites the fewest, while Gemini names more distinct brands per answer than ChatGPT does (Semrush). An engine pulling a dozen or more sources into one answer is going to cite a lot of review sites, news sites and comparison pages that never get named as the recommendation. News sites alone made up an average of 29.5% of ChatGPT's citations among the top 100 UK domains.
A separate Semrush study with Kevin Indig, looking at more than 50,000 brands in ChatGPT, found the same split from a different angle. Only 21% of the most-cited domains in a category were also the most-mentioned brand in that category (Semrush and Growth Memo, July 2026). My guess is the engine reads a third-party page and names the brand the page is about, so the review site gets the citation and the brand gets the mention.
That lines up with what I saw in my own AI Verdict Study. When I sorted the citations from those runs to see where AI engines go to decide about your brand, about four out of five of the ones I could classify pointed somewhere other than the brand being asked about.
Which one matters more
If you made me pick, I'd say mentions are closer to the thing you care about, and I'll explain why, though I'd push back on anyone who drops citations from the report.
In Kevin Indig's earlier Growth Memo research, 74% of users chose the top-mentioned brand as their final pick (cited in the Semrush study). That makes the mention the verdict, the part of the answer the buyer reads and acts on, and it's the part that shows up in someone's shortlist whether they ever click anything or not. I wrote about that no-click problem in The Click That Didn't Happen, and it's the reason I don't think a citation count tells you much on its own.
Citations still matter for two reasons. They're the only part of the answer that can send you a visit (I covered how to find those visits in how to track ChatGPT traffic), and they show you which pages the engine trusts on your topic. If you're cited a lot and mentioned rarely, you're useful to the engine as a reference and it's still recommending somebody else.
How to report AI brand mentions vs citations
What I'd urge you to do with this doesn't need a new tool.
Number one, split the report. Put mention rate and citation rate in separate columns. If your vendor gives you a single "visibility" score, ask them what goes into it, because you need to know whether it's counting your name in the answer, your URL in the sources, or some blend of the two. I went through more of these vendor questions in how to read an AI visibility study.
Two, break it out by engine. The Tesco and Asda numbers looked completely different on ChatGPT and Gemini, and in my own verdict research the same buyer got a different answer in 29% of brand and engine pairings (Same Buyer, Same Brand, Four Different Verdicts). An average across engines hides the one where you have a problem.
Three, figure out which gap you have, because the fix depends on it. If you're mentioned and not cited, the engine has an opinion about you that it formed somewhere else, so go find those pages and check what they say, and give the engine something on your own site worth quoting (clear pricing, clear specs, plain statements about who you're for). The reverse case, cited and not mentioned, means your content is doing the engine's homework while it names a competitor, and I'd read that as a sign your pages explain the category well and say very little about why to pick you.
Then last, watch the sources that aren't you. In the Semrush data, any two platforms shared only 63 to 70 of their top 100 mentioned brands in the UK and 70 to 78 in the US, and agreement on sources was lower still, around 48 to 52 out of 100 in the UK (PPC Land). You don't have to track every one of those domains, just the handful of third-party pages that keep coming up when the engines talk about your category, and what they say about you.
Does your report split the two?
Six questions about how your team tracks AI visibility. Nobody sees your answers but you.
0 of 6 answered
If you want a starting point, the AI Visibility Audit Kit walks you through running these checks yourself, mentions and citations both, across the engines your buyers use. It's $199 and it's built for teams doing this in-house. If you'd rather see where you stand first, the free Quick Check is a lighter version.
A caveat on the data
Semrush says plainly in the report that it doesn't establish that a visibility change caused a commercial result, and I'd take that seriously. These are prompt sets built from search data, run on a schedule, which isn't the same as watching real buyers. Still, it's tens of millions of prompts a month over seven months, and it matches what smaller studies, mine included, have been showing for most of the year.
If you want more on how AI engines decide which brands to recommend in the first place, that's what my book, Explainable, is about.
I'd love to hear what you're seeing. If you've split your own numbers into mentions and citations, send me what the gap looks like, especially if it doesn't match what I've described here.
Jarred Smith is the author of Explainable: Why AI Recommends Some Brands & Ignores Others, an Amazon bestseller on AEO, GEO, and SEO. He's a marketing leader with nearly 20 years of experience across healthcare, public media, retail, and environmental services. Find him at jarredsmith.com.