Semrush, the search data company Adobe bought this year, today published the second 2026 edition of its AI Visibility Index, adding a UK benchmark for the first time. Built on seven months of data from January to July 2026, the 57-page report finds that ChatGPT, Gemini, Google AI Mode and Google AI Overviews agree less with each other in the UK than in the US on which brands to mention and which websites to cite.

In Short

Semrush checked millions of questions put to four AI tools in Britain and America to see which brands the answers named and which websites they used as sources. Being named in an answer and being used as a source turned out to be two different things, and the four tools disagreed with each other more in Britain than in America. A brand that looks strong in American AI answers cannot assume the same picture holds in Britain, or on every AI tool at once.

What Semrush measured

The Index draws on two kinds of collection, and the difference matters for reading every figure that follows. ChatGPT, Gemini and AI Mode each received the same fixed pool of 2.5 million UK prompts and the same pool of 25 million US prompts every month, according to Semrush. Google AI Overviews was measured on a separate keyword-based collection, with monthly bases ranging from 4.35 to 4.80 million in the UK and 48.34 to 54.60 million in the US. Semrush states that these bases are not added into a deduplicated prompt total.

The pools themselves are not alike. The median ChatGPT prompt in the UK was 59 characters long, against 21 to 22 characters for an AI Overviews keyword; in the US the medians were 59 to 60 against 24. The report attributes the gap to how the two collections were built - conversational prompts versus search keywords - rather than to any finding about how people write.

Each platform, market and month produced ranked exports of up to 10,000 mentioned brand domains and 10,000 cited source domains after exclusions. The report sorts results into 22 industry categories, but those categories describe the domain, not the question that produced the answer. A Finance result therefore refers to domains classified as Finance, not to every answer about money.

Two definitions carry the whole report. A mention means an AI answer named a brand, recorded against that brand's domain. A citation means a domain appeared as a source. According to the report, a mention "does not mean the platform recommended it, preferred it, or described it positively," and a citation "is a recorded source position, not a measure of trust or editorial approval."

The UK figures describe one market-level dataset. They are not broken out for England, Scotland, Wales or Northern Ireland, and the methodology states that the Republic of Ireland is not included.

Recurrence rather than growth

The methodology contains a constraint that shapes what the study can and cannot say. Every comparison is made inside a single month, and the report then counts in how many of the seven months a result holds. A range such as "36.7 to 45.3" gives the lowest and highest of those separate monthly results; it is not a trend line. "This edition tests recurrence, not growth," the report states, adding that it "does not isolate platform changes from collection differences" and does not attribute any result to a model update.

That is a narrower claim than much of the AI search measurement market makes. The report also states, in its 90-day planning section, that "this study does not establish that a visibility change caused a commercial result."

How this edition differs from June

PPC Land covered the US-only expansion of the Index when it went live at the end of June. That edition analysed 126 million US prompts collected from January to April 2026 across the same four platforms and 22 categories, and it identified 36 brands that held a top-100 mention position on every platform in every month. Its companion survey of marketers found that 81% of organisations integrating SEO and AI visibility work reported more AI-driven traffic or leads, against 36% of those running the two separately - a self-reported figure.

The October edition extends the window to July and adds the UK. It also changes how some measures are built. The June list, which Semrush called the Universal 36, covered January to April under the exclusions in use at the time. The new Universal Core covers January to July under a revised exclusion policy. According to the methodology, 34 of the June 36 carry over; two dropped out across the longer window and two entered under the new exclusions. "The two lists are related but were built differently, so they are not directly comparable," the report states.

The US count happens to land on 36 again. That number should not be read as stability. It is a different list built over a different window.

Some per-answer figures also look different. The June edition reported mention-to-source overlap of about 64% on AI Overviews and about 30% on Gemini. The new report gives monthly ranges of 44.9 to 63.8 for AI Overviews in the UK and 47 to 62.1 in the US, and 33.4 to 36.9 for Gemini in the UK and 37.7 to 39.4 in the US. The new figures average 22 category-level overlaps between top-100 lists, so the two editions are not measuring the same thing in the same way.

Mentioned is not cited

The central finding repeats from June: being named by an AI platform and being cited as a source are separate positions. Averaged across 22 categories and four platforms, fewer than half of the top 100 mentioned brands also appeared in the matching top 100 citation list, in every month and in both markets. The UK range ran from 36.7 to 45.3 domains out of 100; the US range from 41.4 to 47.4.

The platforms differ sharply on this measure. ChatGPT showed the lowest overlap, at 28.7 to 32.3 in the UK and 32.8 to 35.8 in the US, and stayed below half in all seven months in both markets. Gemini also stayed below half in every month, at 33.4 to 36.9 in the UK. AI Mode dipped below half in five of seven months in each market. AI Overviews, at the other end, fell below half in four of seven UK months and only two of seven US months.

Tesco and Asda

The report opens with two UK grocers. Tesco and Asda both reached the top 100 for mentions on all four platforms in every month measured. Their source positions diverged. In July, Tesco ranked as ChatGPT's 25th most-cited UK domain, while Asda ranked 606th, according to Semrush. On Gemini the two were much closer, at 29th and 60th. Their mention ranks in July were 9 and 8 on ChatGPT, and 24 and 18 on Gemini.

Asda reached Gemini's top 100 cited domains throughout the period but remained outside ChatGPT's top 500. The two brands, in other words, are named at similar rates by both platforms, but only one of them supplies the pages ChatGPT draws on.

The report uses the same grocer to illustrate a second measure, "mutual companions" - two domains that each appear among the other's five closest recorded companions within the same platform and category. Tesco was Asda's mutual companion in all seven months. Sainsbury's and Marks and Spencer qualified in three months each. Morrisons, Aldi and Lidl met the two-way test in none. The report is careful here: "A pairing alone does not establish that two brands compete."

The One-Way Door

The report names a recurring imbalance the One-Way Door: domains inside all four platforms' top-100 mention lists while sitting outside all four top-500 citation lists. The reverse position - heavily cited but rarely mentioned - is much rarer. Across the seven months the UK had 3 to 9 One-Way Door domains and zero in the reverse position; the US had 8 to 13 against 0 to 1.

Seven US domains held the One-Way Door position every month: Chevrolet, Disney, Ford, Honda, Marvel, Nintendo and Toyota. Warner Bros qualified in six. No UK domain qualified in all seven months, although Ford qualified through ford.com in five months and ford.co.uk in the other two, and disney.com also qualified in five.

The imbalance held at wider cutoffs. In a separate July test, Semrush varied the qualifying list from the top 50 to the top 500. At the top 500, 105 UK domains and 136 US domains qualified on mentions while sitting outside the top 500 for citations, against 27 and 32 in the reverse position.

Ford's US numbers show the scale of the gap. In July, ford.com ranked 18th for mentions and 1,055th for citations on ChatGPT, 17th and 1,146th on Gemini, 23rd and 1,320th on AI Mode, and 14th and 748th on AI Overviews. The report pairs this with Toyota: Ford's domain outranked Toyota's for mentions on ChatGPT and AI Overviews in both markets, but on citations, Ford led in the UK while Toyota's domain led in the US.

The reverse case exists in category cuts. In US Finance on ChatGPT in July, Investing.com ranked seventh among cited domains and 50th among mentioned brands. In Industrial, McMaster-Carr ranked eighth as a source and 50th as a mentioned brand.

Only two domains, Apple and YouTube, stayed inside the top 100 for both mentions and citations on every platform in both markets in every month, with no rank gap greater than 25 places. In the US, Reddit and Walmart also met that test every month.

Each platform builds a different answer

The report finds consistent structural differences between the four platforms, which hold across both markets.

Gemini named more distinct brands per answer than ChatGPT in every month, at 3.53 to 5.42 in the UK against 2.75 to 4.18 for ChatGPT. The source comparison ran the other way. ChatGPT carried 12.5 to 22 source entries per answer in the UK and 12.6 to 21.7 in the US; Gemini carried 1.6 to 4.4 and 1.4 to 3.3. AI Mode and AI Overviews sat in between. These counts are URL entries per answer, not unique websites, and include answers that cited nothing.

ChatGPT's link behaviour has been changing during the period. PPC Land reported in April that average unique domains per ChatGPT response fell by roughly 20% after the GPT-5.3 update, based on analysis shared by Lily Ray and Resoneo. The Semrush report does not attribute any change to a model update, so its 12.5-to-22 range cannot be used to confirm or test that observation.

AI Overviews stands apart

Google AI Overviews had the most distinctive mix of brand mentions. In both markets, every pairing that involved AI Overviews shared fewer of its top 100 brands than any pairing between the other three platforms, in all seven months. In the UK, AI Overviews shared 50.6 to 60.4 brands with ChatGPT, while ChatGPT and Gemini shared 78.4 to 82.9.

On sources, AI Overviews and ChatGPT formed the least-overlapping pair in every month in both markets, sharing 42.1 of 100 leading domains in the UK and 43.6 in the US on the seven-month mean. The closest source pair was Gemini and AI Mode, at 57.0 in the UK and 68.9 in the US.

AI Overviews also contributed the most domains that reached only one platform's top-100 mention list: 36 to 49 in the UK and 32 to 47 in the US. By comparison, ChatGPT contributed 7 to 12 in the UK and 9 to 13 in the US.

Concentration on AI Mode

Google's AI Mode, which reached UK users on July 28, 2025, concentrated its citations more heavily than ChatGPT in every month. Within the top-10,000 source export, AI Mode's ten leading domains held 23.2% to 35.8% of citation weight in the UK and 27.9% to 39.2% in the US. ChatGPT's top ten held 12.8% to 15.4% and 14.3% to 15.8%.

The classification of those sources deserves scrutiny. On the seven-month average, Semrush's taxonomy classified "business software and SaaS platforms" as AI Mode's largest source category in both markets, at 17.7% in the UK and 20.1% in the US. The report then states that much of that weight comes from social and community platforms, naming Facebook, Instagram, Pinterest and Quora. A category label covering Facebook and Quora as business software tells a reader less than the domain names themselves.

Where the UK and US part ways

The UK result is consistent across the measures Semrush chose. Any two platforms shared 63 to 70 of their top 100 mentioned brands in the UK on average, against 70 to 78 in the US, according to the press release; the report's chart gives 62.6 to 69.8 and 70.3 to 78.1. Source agreement out of 100 ran from 48.3 to 51.9 in the UK against 54.3 to 56.9 in the US. Shared source positions out of ten were 2.3 to 2.8 in the UK and 2.9 to 3.3 in the US. Meanwhile the UK's "source shelf" was longer: 449 to 484 distinct domains filled category source lists, against 426 to 459 in the US. All four contrasts held in all seven months.

A broader set of UK domains reaches at least one platform's top 100 - 173 to 198, against 159 to 183 in the US. Fewer of them hold on all four: 16.7% to 26.0% in the UK, against 23.5% to 36.5% in the US. Exactly one platform accounted for 38.7% to 47.5% of the UK set and 34.0% to 45.4% of the US set.

A retail-led UK core

The Universal Core - recorded domains present in all four platforms' top-100 mention lists in every month - counted 24 UK records and 36 US records. Fifteen were shared: Adobe, Amazon, Apple, Etsy, Facebook, Google, Instagram, LinkedIn, Microsoft, Netflix, Reddit, Samsung, Spotify, TikTok and YouTube.

Adobe, which owns Semrush, appears on that shared list. Nothing in the report suggests the result was adjusted, but the ownership is relevant context for readers.

Six of the nine UK-only domains were classified as ecommerce or retail: Amazon UK, Argos, Asda, eBay UK, Marks and Spencer and Tesco. The BBC, Nike and TripAdvisor UK completed the group. The 21 US-only domains were broader, spanning automotive (Chevrolet, Ford, Honda, Toyota), sport (MLB, NBA, NFL), entertainment (Disney, Hulu, IMDb, Marvel, NBC, Prime Video, Nintendo) and retail (Best Buy, eBay, Home Depot, Walmart), along with Indeed, Yelp and usda.gov.

Several of those US Universal Core members - Chevrolet, Disney, Ford, Honda, Marvel, Nintendo and Toyota - are the same domains that met the One-Way Door test every month. They are named on all four platforms consistently, and cited on none of them within the top 500.

Just outside the UK core, Boots, Disney, Expedia UK, Ford, Indeed, John Lewis and Sainsbury's qualified in five or six of seven months. Another 35 UK domains qualified in at least one month without making the core.

Because the core counts recorded domain forms, one company can appear twice. Amazon appears as amazon.com in the shared list and amazon.co.uk in the UK-only list.

News carries more weight in UK answers

News and information sites led ChatGPT's source mix in both markets, but more heavily in the UK. They accounted for 29.5% of citation weight among ChatGPT's top 100 UK source domains on the seven-month mean, within a range of 26.4% to 34.3%. In the US the mean was 19.6%. Behind news, the markets split: trip planning (10.1%) and online marketplaces (8.5%) followed in the UK, while online learning (12.3%) and streaming (10%) followed in the US.

On Gemini, news led the UK mix in every month, at 16.6% to 23.0%, followed by online marketplaces and streaming. In the US, streaming led Gemini's mix in five months and online learning in the other two, with seven-month means of 17.2% and 16.0%.

AI Overviews had a different profile. Video streaming sites, with YouTube holding the majority, took the largest share of its citations in both markets - 19.4% in the UK and 24.9% in the US on the seven-month mean. Medical reference sites were the distinguishing feature, at 13.4% to 18.9% of citation weight in the UK and 8.4% to 13.7% in the US; AI Overviews had the highest medical share of the four platforms in both markets, in every month. UK government sources added a further 9.4% on average. The chart labels that UK category "Federal government agencies", which appears to be a taxonomy label carried over from US classification rather than a description of UK institutions.

The medical figure arrives as Google's own health traffic grows. At its Check Up event in March, Google said it handles over one billion health questions a day.

Category matters more than platform

The report argues that category differences outweigh platform differences on the mention-to-source overlap measure. The spread between the highest and lowest categories was larger than the spread between platforms in every US month and in five of seven UK months. In the UK, categories spread from 24.8 to 33.0 points against 14.3 to 31.7 across platforms. In the US the figures were 33.5 to 39.5 against 11.2 to 26.7.

To compare categories across markets, Semrush uses a dual-role share: the proportion of filled top-ten source positions held by domains that also appear on at least one of the category's four top-30 mention lists. Nine of 22 categories had the same market ahead in every month. Twelve categories never reversed direction.

Sports

Sports showed the widest market divide in the report. The UK dual-role share was at least 35 points lower than the US in every month. Four US sports domains - CBS Sports, Fox Sports, MLB and Sports Illustrated - appeared on all four platforms' source lists every month. In the UK, only BikeRadar did.

In July the UK sports source shelf included TalkSport, Golf Monthly, Cycling Weekly, RunRepeat and the Racing Post, none of which reached a top-30 mention list. Manchester United and the Premier League appeared on mention lists but not source lists. Sky Sports, Formula 1, the Olympics and ESPN UK held both roles.

Finance

In UK Finance, Lloyds Bank appeared on all four top-30 mention lists in July while its domain appeared on none of the four top-ten source lists. MoneySavingExpert and Wise appeared in all four UK Finance source lists in every month; the US had no equivalent recurring Finance source. Other UK Finance domains with both roles in July included Uswitch, MoneySuperMarket, MoneyHelper, Compare the Market, GoCompare, NatWest, the Post Office and the London Stock Exchange. The US equivalents included Investopedia, NerdWallet, Bankrate and Morningstar.

The UK dual-role share in Finance was higher in five months, tied in one and lower in one. The report notes that this "describes the lists, not financial performance or the reason a source was selected."

Real estate and industrial

Real estate showed recurring source cores in both markets. In the US, Apartments.com, Realtor.com, Redfin and Zillow appeared across all four source lists every month. The UK had OnTheMarket, Rightmove and Zoopla. The US dual-role share was higher throughout, by 7.5 to 30 points.

Industrial had no recurring US core at all, and its US dual-role share was lower than the UK's in every month by 10 to 35 points. Food and Drinks had no recurring all-platform source domains in either market, nor did Beauty in the UK.

Vendor material and the claims attached to it

The Index is a Semrush publication, and the report doubles as a product brief. Its planning section describes running ten customer questions across four platforms as an "afternoon AI visibility audit" and proposes a 90-day cycle in which a team sets baselines, chooses an intervention and then decides whether to expand, revise or stop. It points to "Semrush's Enterprise AI Optimization" for prompt generation and daily tracking and to Semrush's "AI Operating System" at a separate sign-up address. The landing page gates the report behind a form requesting name, business email, country, phone number and company size, and describes it as a "Flagship Report" published in October 2026 for "cross-functional teams".

The technical checklist in that section is more specific than the product mentions. It lists robots.txt, indexing settings and server or CDN restrictions as items to review, and notes that "a chatbot's statement that it can reach a page is not a technical test." A table assigns gaps to teams: inaccessible pages to SEO and engineering, outdated facts to content specialists, inaccurate third-party descriptions to communications.

The commercial figures in the release rest on Adobe data that the report does not detail. Both documents state that AI-referred visitors "convert 60% better than non-AI traffic and generate 53% more revenue per visit." Neither gives a time period, a sample, the sectors covered or how conversion was defined. When Adobe made its LLM Optimizer generally available in October 2025, it reported that generative AI visitors converted at rates 5% higher than traditional channels, based on September 2025 retail data. A move from 5% to 60% may reflect a different period or sample, but the documents do not say. Microsoft Clarity data covered by PPC Land in November 2025 found AI referrals converted at higher rates than search while making up under 1% of traffic across the domains analysed.

There is also a small discrepancy in titles. The press release identifies Andrew Warden as "Global Vice President of Marketing and CMO of Semrush, An Adobe Company"; the report's foreword gives "VP Marketing and CMO of Semrush, An Adobe Company".

What Adobe and Semrush said

Marissa Dacay, Global Vice President, Enterprise Marketing at Adobe, said in the release: "For enterprise brands, brand and AI visibility is a global concern. And with every additional market comes additional complexity. Brands need to optimize their own content for the audiences, questions, and expectations of each market, while also building the right third-party signals around them."

Warden described AI as "now a decision layer for customers, changing how they discover brands, sort through options and ultimately purchase. That makes visibility a business and commercial growth issue, not only a search issue. The sources and brands that shape decisions in the UK are not identical to those in the US, and teams need to build for those differences while keeping brands consistently discovered, evaluated and chosen wherever AI is shaping the decision."

The release was issued from London, where Semrush presented the findings at its Spotlight London event.

Why it matters for marketers

The Index arrives at a point where AI visibility is a crowded measurement category with no agreed standard. When the IAB published its "Measuring Visibility in the AI Era" guidelines in August, it found that only 16% of brands systematically track AI visibility, and noted that more than 20 vendors sell tools that often disagree when measuring the same brand. The Semrush report does not reference the IAB framework.

Semrush's position in that market has changed during 2026. Adobe agreed to buy Semrush for $1.9 billion in November 2025, at $12 per share in cash, and Adobe said the deal completed in April 2026. Since then Semrush has embedded its data in Lovable's app builder and shipped a Claude connector covering 28.8 billion keywords and 317 million LLM prompts. Rival platforms are consolidating too: Zeta Global agreed this month to buy Senso, the startup whose technology powers its GEO product.

The UK angle has particular weight. Recrawled, a UK agency, found earlier this month that ChatGPT, Gemini and Google never named 9 of 40 UK software firms in shortlist answers, and that adding or dropping "UK" from the question moved the share of firms named from 63% to 43%. That was a one-day test on 40 companies. The Semrush data covers seven months and millions of prompts, and it points in the same direction: UK results differ from US results, and they differ between platforms.

UK publishers have their own interest in the news finding. If news sites account for nearly 30% of ChatGPT's top-100 citation weight in the UK, publishers are a large part of the source layer AI answers depend on - at a time when UK sites have reported organic traffic growth falling sharply since Google's AI search rollout, and the CMA has pushed Google to explore opt-out controls for AI Overviews and AI Mode. The report counts citations; it does not measure whether those citations send traffic.

Platform choice carries weight as well. Comscore data reported in September showed ChatGPT's share of US AI prompt volume falling from 70% to 50% between January and June 2026, while Gemini rose from 17% to 30%. The Semrush study does not weight its findings by platform usage; it treats the four platforms as parallel environments. Claude, Perplexity and Copilot are not included.

The mention-versus-citation split also aligns with smaller independent work. Suganthan Mohanadasan's analysis of raw ChatGPT responses, published in August, found that brands named in ChatGPT's own search query were mentioned about 33 times more often than brands that were only retrieved, and that 3.1% of retrieved pages were cited. That study was based on one account. It does, however, offer a mechanism for how a brand could be named often while its pages are rarely cited - the pattern Semrush describes as the One-Way Door.

What the Index does not answer is whether any of these positions affect sales. The report is explicit on that point, and its recurrence-based method rules out claims about movement over time. Its value lies in showing, across a dataset larger than most independent studies, that a single AI visibility score hides at least four separate measurements - by platform, by market, by category, and by whether a brand is named or cited.

Timeline

Summary

Who: Semrush, the brand visibility platform owned by Adobe, with comments from Andrew Warden, CMO of Semrush, and Marissa Dacay, Global Vice President, Enterprise Marketing at Adobe. The data covers brands and publishers including Tesco, Asda, Ford, Toyota, Lloyds Bank, MoneySavingExpert and the BBC.

What: The H2 edition of the 2026 AI Visibility Index, a 57-page report comparing brand mentions and website citations across ChatGPT, Gemini, Google AI Mode and Google AI Overviews in the UK and US. It finds the four platforms agree less in the UK, that fewer than half of top-mentioned brands are also top-cited sources, and that news publishers carry more citation weight in UK answers.

When: Published today, October 13, 2026, covering seven monthly measurements from January to July 2026.

Where: Presented at Spotlight London; the data covers one UK market-level dataset and the US, across 22 industry categories.

Why: Semrush says brands cannot rely on a single global AI visibility ranking because results differ by platform, market and category. The study measures recurrence, not growth, and states that it does not establish any link between visibility changes and commercial results.