Every vet practising in Britain is on a public register. You cannot call yourself one without it. It is the most verifiable credential in local services, far cleaner than “trusted”, “award-winning” or any of the other things a website says about itself.
We scored 1,977 UK veterinary practice websites for how ready they are to be recommended by AI. Only 12% name a qualified vet in a way a machine can read. The other 88% are, to the systems now doing the recommending, anonymous.
Not one practice in the country is AI-ready.
They did the hard part first
This is the bit I did not expect. Vets build good websites. On technical basics they average 79 out of 100, which is near the top of every sector we have measured. Sites load, they work on a phone, they are secure, the plumbing is sound.
Then look at what AI actually leans on.
| Area | Average / 100 |
|---|---|
| Technical basics | 79 |
| Easy for AI to quote | 72 |
| Topic coverage | 64 |
| Platform setup | 55 |
| Expertise and trust | 46 |
| AI-platform readiness | 45 |
| Labelled information | 36 |
| Brand authority | 34 |
Seventy-nine at one end, thirty-four at the other. The expensive, technical work is done. The cheap, descriptive work is not.
Key Takeaways
- They did the hard part first
- What “AI can read it” actually means
- The two fixes almost nobody has made
- Seventeen practices are a weekend away
That ordering matters more than the average score, because it tells you what the constraint actually is. It is not budget, and it is not competence. A practice that can commission a fast, secure, mobile-ready website can certainly write down which of its vets is a vet.
What “AI can read it” actually means
A page that says “Meet our team: Sarah, Tom and Priya” is legible to you and close to meaningless to a machine. It does not say that Sarah is a veterinary surgeon, that she holds a registrable qualification, or that this is a claim about a person rather than a caption on a photograph.
The same information, marked up so a machine can parse it, becomes a fact a system can repeat with some confidence. That is the entire difference, and it is a plugin and an afternoon.
Practices that do it score around 12 points higher.
The two fixes almost nobody has made
Structured FAQ markup and AI-answer markup are each worth around 12 points. Fewer than one practice in fifty has done either.
I want to be careful about how much weight that carries, because those checks are inputs to our own score, so some of that 12 points is arithmetic rather than the world. What the arithmetic does not explain is the distribution. If these were hard, you would expect the best-resourced practices to have done them and the rest not to have. Instead almost nobody has, across every band. That is not a resourcing pattern. It is a nobody-mentioned-it pattern.
Seventeen practices are a weekend away
Seventeen practices score between 70 and 80. The bar for AI-Ready is 80. Four in five of them are missing the same single thing: structured FAQ markup.
The closest is Your Family Vets at 78.1. Nobody in the country has cleared 80.
So the first AI-ready veterinary practice in Britain is currently available, and the work involved is an afternoon rather than a rebuild. I find that a more interesting fact than the average.
Why this sector, specifically
Vets sit mid-pack on our measure: around 52, level with dentists and accountants, behind aesthetic clinics at around 56, ahead of law firms at around 47. On the raw number this is an unremarkable sector.
What makes it worth writing about is the shape. Veterinary medicine is a profession whose entire product is a credentialed human making a judgement about an animal you love. The credential is real, mandatory and publicly verifiable. And in seven cases out of eight, it is not on the website in a form the recommending systems can use.
When an owner asks an assistant to recommend a vet near them, it names two or three. Not ten links, two or three names. The practices that get named are not the better clinicians. They are the ones whose expertise was written down in a way a machine could read.
Method, and what this does not show
The figures here come from the published SearchScore SAVI report for UK veterinary practices 2026: 1,977 practice websites, scored 0 to 100 across eight areas. I have not recalculated anything for this piece. The report is the benchmark; this is my reading of it.
Disclosure: I built SearchScore, so this is my own instrument and I have an interest in you finding it useful. The method is published at searchscore.io/methodology, and where our numbers are not yet good enough for a given decision is published at measurement standards.
What this measures: readiness, not visibility. The score reads what a website publishes and predicts whether AI systems can find, understand and cite it. It does not observe whether ChatGPT actually recommended a given practice last Tuesday. Those are two different instruments and I try not to blur them.
What it cannot see: quality of care. Nothing here is a comment on clinical standards, and a low score is not a bad vet. If anything the point is the reverse: the care is not the variable, the labelling is.
If you run a practice
Three things, in order, none of which need a new website.
- Name your vets, with their qualifications, marked up so a machine can read it rather than only a visitor.
- Take the questions owners actually ask you on the phone, answer them on the site, and tag them as questions and answers.
- Then check whether anything changed, rather than assuming.
You can score your own site in about sixty seconds at searchscore.io, no email required. More of my reading of this data is in the research, and the underlying discipline is set out in the guide to generative engine optimisation.
The vets did the hard part. Then 88% skipped the easy one.
