Tesco and Asda were among the 100 brands mentioned most often in UK answers on ChatGPT, Gemini, Google AI Mode and Google AI Overviews in every month from January to July 2026. In July, ChatGPT ranked Tesco’s website its 25th most-cited UK source. Asda’s ranked 606th.

That gap sits at the centre of the latest edition of Semrush’s 2026 AI Visibility Index, which the Adobe-owned company released at its Spotlight London conference on 13 October. The study separates two things a single AI visibility score can blend together: whether an AI answer names a brand, and whether it uses that brand’s own website as evidence. The distinction matters because the two positions point to different work. A brand that is named but not cited has a source problem, often decided by publishers, retailers and forums. A brand that is cited but rarely named has a recommendation problem. A single composite score cannot say which one a team has.

What did Semrush’s AI Visibility Index find about mentions and citations?

Semrush found that, averaged across 22 categories and four AI platforms, fewer than half of the top 100 mentioned brands also appeared among the top 100 cited domains, in every month measured and in both the UK and US. In the UK, the average overlap ran from 36.7 to 45.3 domains out of 100, depending on the month.

The figures come from fixed monthly sets of 2.5 million UK prompts and 25 million US prompts, built from Semrush search data and run through ChatGPT, Gemini and AI Mode, with a separate, larger set of AI Overviews results for real search keywords. This is the first edition to include a UK benchmark. Semrush’s earlier edition, published in June, covered US prompts only.

The overlap measures rankings, not individual answers. It does not mean fewer than half of AI responses cite a brand’s own site. It means the list of brands AI talks about most and the list of websites it leans on most are, for the most part, different lists.

The US examples run wider. MarTech, which Semrush owns, reported from the full study that Ford appeared in all four platforms’ top 100 mentions in July while its domain ranked 1,055th for citations on ChatGPT and 1,146th on Gemini. Chevrolet, Disney, Honda, Nintendo and Toyota were also named often everywhere, yet sat outside the top 500 cited domains in every month. The pattern runs the other way too: in US finance on ChatGPT, Investing.com ranked seventh among cited domains but 50th among mentioned brands.

Why can a single AI visibility score mislead marketing teams?

A single AI visibility score averages across platforms and categories that, on Semrush’s data, behave very differently. Gemini named more distinct brands per answer than ChatGPT, while ChatGPT cited the most sources per answer and Gemini the fewest. Google AI Overviews had the lowest brand-mention overlap with the other three platforms in every month, and its pairing with ChatGPT shared the fewest sources of any two platforms.

The UK is more fragmented than the US. Any two platforms shared 63 to 70 of their top 100 mentioned brands in the UK, against 70 to 78 in the US. UK answers also lean harder on news: news sites took an average 29.5% of ChatGPT’s citations among its 100 most-cited UK domains, compared with 19.6% in the US.

Category adds a further layer. Semrush found that differences between categories exceeded differences between platforms on mention-to-citation overlap in every month. US real estate drew on the same four source domains on all four platforms every month (Apartments.com, Realtor.com, Redfin and Zillow), while US industrial had no equivalent recurring set.

For a B2B marketing team, that turns one question into several. A brand can be named on ChatGPT and missing from AI Overviews, or strong in US answers and weaker in UK ones. Andrew Warden, Semrush’s chief marketing officer, described AI visibility as “a business and commercial growth issue, not only a search issue”. The practical consequence is narrower: a team cannot tell whether a weak score calls for new content, digital PR or work with third-party sites until the mention and citation figures are pulled apart. MarTech’s analysis of the study argues that where a brand is named but its site is not cited, the fix may belong with communications or content teams rather than SEO.

How far can marketers rely on the Semrush figures?

The Semrush figures come from a vendor that sells AI visibility tracking, and most of the coverage so far has come from inside its orbit. MarTech, which published the most detailed breakdown, discloses that it is owned by Semrush. The Drum’s coverage was sponsored content produced with Semrush. The Next Web’s report quoted only Semrush. None of that makes the data wrong, but no independent analyst has yet audited the prompt sampling or how each platform’s citations were parsed.

Semrush’s public materials also leave gaps. The announcement describes 22 categories without listing them, and the category-level overlap tables sit in the downloadable report, which MarTech says requires registration. No B2B software category result is public, so the headline figures should not be read as proof of a gap in any one sector.

Separate data from BrightEdge, another tracking vendor, points the same way on the core distinction. Its study of ecommerce prompts across ChatGPT, Gemini and AI Overviews from 10 May to 26 July 2026, published on 30 July, found that brands’ share of mentions moved by an average of 25% a week on both ChatGPT and Gemini, while share of citations moved by 39% on ChatGPT and 41% on Gemini. Only 13 of the top 50 cited domains were shared across all three engines, against 28 of the top 50 mentioned brands. BrightEdge measured a different sector over different weeks, so it is not a replication, but it suggests citations are both less shared and less stable than mentions.

Semrush’s report does not show that changes in AI visibility caused any commercial result, and MarTech’s own write-up says as much. The full study, 2026 AI Visibility Index: Where the Top Brands Stand in the UK and US, is available from Semrush’s index site, with the category tables and methodology behind its download form.