UK GP Practices
Search every site in this cohortuk-gp-practices
Every active general practice in the four nations.
Middle of 31 cohorts. All cohorts 54 · best cohort 63.
The source, and how we filtered it
| Source | GP practices in the UK, NHS England (Organisation Data Service; NHS.UK practice profiles) (2026-11-02) |
|---|---|
| License | ODS: Open Government Licence v3.0. NHS.UK practice profiles: Open Government Licence v3.0 under NHS.UK's terms (sections 3.4 to 3.11), which permit any use including commercial. Only each practice's name and website address are taken, which NHS.UK's terms treat as adapted content, so it is attributed with the generic OGL statement and not as information from the NHS website. Read 2026-11-05. |
| Cite as | Contains public sector information licensed under the Open Government Licence v3.0. |
| How we filtered it | Active organizations with the ODS GP practice role (RO76) for England, Wales and Scotland, and active primary role RO315 for Northern Ireland. For each English practice, the website link on its NHS.UK profile (6,256 pages, read one at a time). Addresses are cut to the root unless several practices share a host with a page each. |
| Instead of | The NHS.UK Service Search API (needs a key); the PHS, NHS Wales and BSO practice lists (no websites). |
| Built | 2026-11-02 |
The filtering is ours and is disclosed as ours. The source published a list; the decision about which rows of it belong in a cohort was a judgment call, and it is written down so somebody can disagree with it.
Where the 4399 sites land
| Grade | Share of the cohort | Sites | Share |
|---|---|---|---|
| Adequate | 1471 | 33.4% | |
| Weak | 2511 | 57.1% | |
| Not ready | 403 | 9.2% | |
| Strong | 14 | 0.3% |
What the cohort earns, dimension by dimension
The mean share of each dimension’s available points that this cohort earned. Low here means the whole sector is weak on it, not that one site dragged an average down. The per-site spread is in the dataset.
| Dimension | Mean share earned | Mean |
|---|---|---|
| #1 · Reach
These parameters determine the fundamental ease of access that AI has to your site information. |
68% | |
| #2 · Read
AI generally turns off JavaScript. Does your site information survive without it? |
56% | |
| #3 · Understand
Does a machine know who and what you are? |
25% | |
| #4 · Act
Sitemaps, redirects, 404s, licensing terms, and not wasting the crawl budget. |
76% |
The cap is 32 URLs and it is a ceiling, not a promise. A portal whose every link points at a different host yields one page; a university yields the full sample. The median here is 32, the range 1–32, published so that "we sampled 32 URLs per site" cannot become a claim the data does not support.
36 sites in this cohort served a certificate we could not verify. They were read anyway and scored normally, because reporting no response for a site every browser loads would describe our TLS stack rather than the site, and each one carries the finding in its row.
11 tests almost nobody in this cohort passes
| Test | What it asks | Failing | Of | Points |
|---|---|---|---|---|
| ACT-3 | If you have a public API, is it described in a standard file? | 1 | 2 | |
| EXT-11 | Have you marked the part of a page that answers a spoken question? | 4399 | 0 | |
| EXT-12 | Do you tell search engines the moment a page changes? | 4399 | 0 | |
| EXT-14 | Does your site offer its actions to AI browsers as tools? | 4031 | 0 | |
| EXT-15 | Can an AI agent find out how to sign in to your site? | 1 | 0 | |
| EXT-16 | Do you publish a card or a skills list for the agents you run? | 4399 | 0 | |
| EXT-17 | Do you list your agents in your domain's DNS records? | 4386 | 0 | |
| EXT-2 | Do you list your APIs and agent tools where agents look for them? | 886 | 0 | |
| EXT-7 | If you run real question-and-answer pages, are they marked as such? | 17 | 0 | |
| RED-10 | Do you offer a plain-text (Markdown) copy of your pages? | 3394 | 2 | |
| UND-9 | If a page is behind a paywall, does it say so? | 1 | 2 |
| Test | What it asks | Failing | Of | Points |
|---|---|---|---|---|
| RED-12 | If a machine asks for Markdown, does your site give it? | 4399 | 2 | |
| UND-1 | Does your site say clearly who you are? | 4397 | 2 | |
| UND-10 | Does your page code link to your official profiles elsewhere? | 4397 | 2 | |
| UND-2 | Does your site use the page labels that search engines still read for what you publish? | 4399 | 3 | |
| UND-5 | Does the author label in your page code point to that person? | 3268 | 1 | |
| EXT-1 | Do you publish an llms.txt file for coding assistants? | 4399 | 0 | |
| PST-2 | Have you posted a Content Signals statement about what AI may do with your content? | 4399 | 0 | |
| EXT-13 | Can an assistant ask your site a question directly? | 4399 | 0 | |
| PST-4 | Have you reserved your text and data mining rights in a form machines can read? | 4399 | 0 |
Coral bars mark tests worth 2 points or more, the failures that actually move a score.
Every 7,901, named, with a report for each
Whether or not it has been scanned. Rows with no website on record, sites that were unreachable and sites that turned us away stay in the table with the reason on them, and none of them enters an average: dropping them would shrink the population to the part that was easy to measure. A few points between two neighbors is noise. The box filters every row in the cohort, not only the page you are on; click a header to sort the rows on this page by name or by score.
Showing 7,901–7,901 of 7,901 · page 80 of 80
| Site | Score | Gated | Pages read | Report |
|---|---|---|---|---|
| Ystwyth Medical Group no website on recordregister id ODS W92025; no published directory carrying Wales's practice websites could be read, and the practice's ODS record gives none | — | No | — |
Scores are for the population, not for ranking two named sites against each other. A few points between neighbors is noise. See what these numbers can and cannot support. Headline figures: summary.json (CC BY 4.0). The site-by-site data on this page is licensed; a bulk copy is from TEN7 for a fee.