The UK cut of 204 readings: whito, rank4ai and aeoscore were cited dozens of times, and we were cited once

In the United Kingdom cut of our 15 October 2026 benchmark, ChatGPT and Gemini each answered 102 buying questions, 204 UK readings in total, and three small British agency sites were cited dozens of times each: whito.co.uk 46, rank4ai.co.uk 39 and aeoscore.co.uk 32. Our own domain was cited once in those 204 readings, on ChatGPT, and not at all on Gemini. The finding for a British publisher is in that gap. The engines were not rewarding size in the UK set. They were rewarding being specifically British about a British question.

This page pulls the British readings out of a larger exercise. The other two markets sit alongside them in what AI assistants actually cite across 597 readings of ChatGPT and Gemini in three markets, where the method and the full source mix are set out.

The UK readings, and who won them

Date run	15 October 2026
Engines	ChatGPT and Gemini
Questions	102, of which 72 are worded identically across the three markets and 30 are UK specific
UK readings	204
Readings citing aiknowsus.com	1 on ChatGPT, 0 on Gemini
UK domains cited most	whito.co.uk 46, rank4ai.co.uk 39, aeoscore.co.uk 32

The market was fixed in the prompt text rather than by a location parameter, and then checked on the answers: the UK prompt came back answered for London, in pounds. That check matters more than it sounds, because the engine surfaces we read accept a country parameter, return success, and ignore it.

Being specifically British beat being large

whito.co.uk, rank4ai.co.uk and aeoscore.co.uk are small agencies. None of them has the reach of the global marketing platforms that lost those same readings, and all three were cited dozens of times while our domain was cited once. Across the whole run, 85.6% of the 10,652 citations recorded went to a page on the vendor's or the agency's own website, so these are not third party endorsements of those firms. They are their own pages, read and used.

That is the part a British business can act on. The engines reached for a .co.uk page about a British version of the question, by a company nobody would call big. There is no size requirement in that result.

What makes a British page the obvious one to cite

• Price in pounds, and state the VAT position. Every build spec conversation in the run came back wanting a price in the reader's currency. A monthly figure with the VAT treatment spelled out answers what a British buyer is really asking.
• Keep the claim inside what you can evidence. UK marketing claims sit under the CAP Code and the ASA can be asked to look at them, so a comparative or superlative claim on a British page needs the evidence attached to it. A dated measurement is the cheapest form that evidence takes.
• Use the British version of the question. Terms, regulators, invoicing, procurement habits and search vocabulary differ, and the engines treated the UK questions as UK questions. The same pattern shows up in ordinary search, which is covered in what actually differs for UK search.
• Name the engines and date the page. 98 of 102 specs asked for the engines to be enumerated rather than described as AI in general, and 102 of 102 asked for a date.

British public sector sources appear in the UK set

The UK readings reach for British institutional sources that the American readings did not produce, gov.uk and ico.org.uk among them. For anything touching data, consent, complaints or compliance, a British page that cites the relevant government or regulator page, and says which one, is lining itself up next to the sources the engines already trust in this market. Those two counts are now published, so here they are: gov.uk was cited in 57 of these readings and ico.org.uk in 28, and neither of them appeared anywhere in the United States set. When this page first went up we held both figures back, because the only per domain numbers we were prepared to stand behind were the ones on the benchmark page. That reservation no longer applies. The same tested classifier produced these two, and pages published on 16 October 2026 carry them, so withholding them here would have left this page refusing a number a reader can find two clicks away.

The parts of the UK set that are not winnable

The questions about how the engines themselves work went to the platforms that build them: developers.google.com was cited in 178 readings across the run, support.google.com in 138, blog.google in 125 and help.openai.com in 112. A British agency page explaining how Gemini picks sources is competing with Google's own documentation. Better to be the layer above it, which is the British application of the thing.

Review directories are the other dead end, at 0.8% of all citations recorded in the run. A UK firm deciding between a year of directory profile building and six well evidenced British pages has a measured answer to that question now.

One scope point, because it changes what this page is worth. Two engines were in this run, ChatGPT and Gemini. The run did not include Google AI Overviews, Perplexity or Claude, and our product covers all five, so these UK counts describe two surfaces out of the five.

What British publishers ask about this result

Why did three small UK agencies beat far larger companies?

On this evidence, because their pages were specifically about the British version of the question. The citation counts are 46, 39 and 32 readings, and the pages cited were their own.

Where did our single British citation come from?

From ChatGPT, one reading out of 204. Gemini cited us in no market in this run, the United Kingdom included.

Does London get measured separately from the rest of the UK?

The UK prompts were locked to the United Kingdom and London, and the answers came back for London in pounds. A city by city British breakdown was not in this dataset when this page was written, and now it is: on 15 and 16 October 2026 we put six local buying questions to ChatGPT and to Gemini in London, Manchester and Birmingham, 12 readings each and 36 in all, and our own domain was cited in none of them. The three cities did not agree with each other either, which is the reason to read them separately rather than as one British figure.

Is a .co.uk domain necessary?

The three most cited UK agency domains in this run are .co.uk. What they share with any winnable page is British specificity in the content, and a UK facing page on a .com can carry that too.

How often is the UK cut re-measured?

At 60 day intervals, on the same frozen set, with the date and the sample size printed on each round.

Where a British page can win a citation

Take one buying question a UK customer asks, answer it with British numbers, in pounds, with the VAT position stated and the date on the page, and name ChatGPT and Gemini rather than AI in general if the subject is AI visibility at all. That is what the three agencies cited 46, 39 and 32 times were doing, and the same reasoning applies to reading a London rate card: the figures in how to read a London rate card for the overhead instead of the output are the kind of specific British detail these engines kept citing.

We should say plainly that AI Knows Us sells this measurement and the publishing work after it, so the UK result above is being reported by an interested party who came out of its own run with a single citation. The method is published, the counts are published, and the next run is dated.