Which of these SEO agencies can actually do AI search?

It is the second-heaviest criterion in the study, behind proven outcomes and ahead of everything else, because it is the question buyers have started asking and almost nobody is answering with evidence.

Nearly every agency here now sells something called GEO, AIO or AISearch. Selling it and being able to do it are different, and the gap between them is what this page is for.

The AI search score is the agency’s place in bestgeoagencies.co.uk, a dedicated study of 20 GEO agencies that reads each one against seven criteria of its own and a tracked prompt set. It is a deeper piece of work on this single question than a side-criterion here could ever be, so it is used rather than reinvented.

11 of the twenty agencies on this page appear in it. Their score there is their score here, carried across unchanged, and the table shows where each one placed. The full GEO ranking, with its own working, is published there.

That study is a sister publication under common ownership with this one, so the eleven carried scores are not independently sourced. It is stated here, beside the score, rather than left in a footer for a reader to find.

It counts how often an agency’s own website is quoted by AI assistants. It is objective, anyone can reproduce it, and it is the closest thing to a neutral measure that exists.

It measures their visibility. Skill is a different thing and this does not reach it. Screaming Frog tops this column with 1527 citations and does not top the ranking, because it runs one of the most linked-to blogs in SEO. That is a fact about their publishing rather than about their AI-search practice. A small agency doing excellent AI-search work for clients and writing none of it up will sit near the bottom.

So it is one input of four and it decides nothing on its own. The others are whether they sell a named service in their own words, whether they publish original work about AI search, and whether any client result with a figure exists.

Carried. 11 agencies were read into bestgeoagencies.co.uk against its seven criteria and its tracked prompt set. Their place there is their score here. The table names the position so a reader can go and read the working.

Scored here. The other 9 are not in that study. They are assessed against the same four measures it weighs: a named service in their own words, published research about AI search, a client result carrying an AI-search figure, and their own visibility in AI answers.

What the second route does not do is run the tracked prompt set. That is the one thing it cannot reproduce, and it is why the two routes are labelled separately rather than blended into a single column a reader cannot take apart.