I publish original research on whether AI search engines can find, understand and recommend businesses. Not opinion about where AI search is heading, but measurements of where it actually is, across whole professions, with the sample sizes and the method stated.
The benchmarks themselves are run through SearchScore, which I built. What is below is my reading of them: what the numbers mean, what causes them, and what I would do about it. Disclosure noted so you can weigh it.
Published
The most AI-visible care home in Britain is rated “Requires improvement”
2,307 homes. Not one is AI-ready, the least visible sector measured so far. A CQC Outstanding rating moves the score by four points: Outstanding homes average 50.0 against 46.1 for Requires improvement. The single most AI-visible home in the country is rated Requires improvement and outscores all 125 Outstanding homes.
One in ten UK law firms built a website AI cannot read
9,487 firms. Three are AI-ready, the thinnest rate in any benchmark so far. The 964 firms with JavaScript-built websites are 76 times less likely to be visible, and account for almost half of every effectively invisible firm in the country. Not one firm in 9,487 reaches the Strong band without publishing structured data.
Zero out of 1,038 accountancy firms write for AI
1,038 firms. One is AI-ready. None lays its answers out in the format AI quotes from, and 82% of the firms that wrote a questions-and-answers section never added the markup that lets a machine read it. The hard part was done and the last step was skipped.
What these have in common
Three professions, wildly different in size, and the same shape every time. Technical basics score in the seventies. Structured data and brand authority score in the thirties. The crawlers get in and there is nothing labelled for them to lift.
Across the editions published so far, the AI-Ready rate runs at roughly one firm in a thousand to one in three thousand. That is not an industry problem. It is what happens everywhere when the loss is invisible: no dashboard reports the client who never got in touch because an assistant recommended somebody else.
Method, in short
Each benchmark checks live websites the way an AI reader would, across more than 250 signals, and scores each business 0 to 100 on whether AI can find, understand and recommend it. Bands are AI-Ready 80 to 100, Strong 60 to 79, Emerging 40 to 59, Low Visibility 20 to 39, Invisible 0 to 19.
Where I quote a benchmark I use its published figures rather than recalculating, and I say so. Where I have derived something myself, I state the denominator and what was excluded. If a number here is wrong I would rather hear about it than not, so tell me.
What is coming
More editions exist and will get the same treatment: aesthetic clinics, veterinary practices, dentists, and the regional cuts. If there is a profession you would find useful, say so.
If you want the underlying method rather than the findings, start with the guide to generative engine optimisation. If you want your own numbers, the audit is free at SearchScore.
The method behind this research is set out at length in my book, Get Recommended by AI.
The vets did the hard part. Then 88% skipped the easy one. 1,977 UK veterinary practices, and not one is AI-ready.
