Think of nhs.uk and you probably picture the health encyclopedia. The plain page that tells you what a rash might be, whether that headache warrants a call, how long a course of antibiotics should run. It is the closest thing Britain has to a national medical reference, and it is what the site is known for.
That is not what most of the site is.
We pulled the full sitemap on 29 August 2026 and counted every entry in it. The health-information pages are there, and they are excellent. They are also a rounding error next to the rest.
The overwhelming majority of nhs.uk is a listing of physical places: clinics, pharmacies, GP surgeries, dental practices. Address, opening hours, phone number, a link to book. Not advice about health, but directions to it.
Once you see that, everything else about the site — how it earns links, how it is built, why it ranks where it does — makes a different kind of sense.
Results at a glance
nhs.uk lists 51,900 provider profiles in its sitemap and 4,037 pages of health information. The thing the site is famous for is the smallest meaningful part of it. The thing nobody thinks about — the directory — is nine tenths of the index.
That directory is also why the site pulls links at a rate almost nothing else at its scale manages. Local pages get cited locally, one practice and one pharmacy at a time, and 51,900 of them add up.
What the sitemap actually holds
The structure is refreshingly simple. One sitemap index, 8 children, none of them nested. Every child is a plain XML file served uncompressed, and the whole tree comes to well under 10 MB. There is nothing to reverse-engineer and no crawl trap to fall into. You can read the entire shape of the site in a couple of minutes.
What the shape shows is a site split by content type rather than by volume. The files are wildly uneven — the largest is more than 500 times the size of the smallest — because each one holds a category, not a chunk. Four of the 8 hold provider profiles. One holds the encyclopedia. The rest hold campaigns, video markup and a handful of search landing pages.
The encyclopedia is the small part
The health-information file holds 4,037 pages, spread across 30 top-level topics. Medicines is the largest of them, conditions is second, and after that the tail drops away fast into tests, mental health, pregnancy and the rest. It is a tightly edited reference, not a content farm, and the restraint is obvious from the counts alone.
Against that sit 51,900 provider profiles. The ratio is close to 13 to 1. For every page explaining what something is, there are roughly 13 pages telling you where to go about it.
That is a deliberate answer to a question most publishers never ask themselves: what does the reader actually need next? Someone who has just read about a repeat prescription needs a pharmacy, not another article. Someone worried about a tooth needs a dentist with an open list. The encyclopedia exists to hand people off to the directory, and the directory is sized for the handoff.
A directory makes a quiet promise
A page that says a pharmacy exists is making a claim about the physical world, and the physical world moves. Practices close, merge, change hands, move borough. An encyclopedia page about a condition can sit untouched for years and stay true. A profile page cannot.
We sampled the profile sets to see how that promise was holding. Three of the four came back clean — every sampled service-directory, GP and dentist page loaded normally. The pharmacy set did not. Roughly an eighth of the pharmacy URLs in the sitemap no longer resolve, with the dead ones clustering in a pattern that suggests providers who now belong to a different part of the health service and were never removed from this one.
This is worth stating carefully, because it is a sample and not a census: the pharmacy count in the sitemap is a listing count, not a live-page count, and the live number is meaningfully smaller. It does not change the shape of the site. It does change what you should trust a raw sitemap number to mean — including on your own site.
It is also the single most copyable lesson here. Nobody notices a stale directory entry from the inside. The sitemap keeps declaring it, the crawler keeps requesting it, and the only signal is a quiet stream of errors that no dashboard is watching.
Why every listing earns its links
Here is where the directory stops being a maintenance burden and starts being the whole strategy.
We ran nhs.uk through the UNmiss Backlink Analyzer and divided its referring domains by its sitemap entry count. The result is 5.2 referring domains per page. In this series that is extraordinary. Only calculator.net and Healthline sit above it, and both of those are far smaller sites where the arithmetic is easier. Every large site we have measured — MDN, GOV.UK, PyPI, Eventbrite, Meetup, Rotten Tomatoes — comes in lower, most of them by an order of magnitude or more.
The comparison that really lands is GOV.UK, another vast British public-sector site measured with the same tool earlier in this series. GOV.UK earns well under 1 referring domain per page. nhs.uk earns nearly 6 times that rate.
The reason is local. A GOV.UK guidance page is national, so it gets cited nationally, by a limited number of national outlets. An nhs.uk pharmacy profile is a specific building on a specific street, and the people who link to it are a parish newsletter, a local charity, a council page, a community Facebook group posting the address of the late-opening chemist. Those links are individually tiny and collectively enormous, because there are 51,900 different listings to link to.
Built plainly, and it shows
The UNmiss Website Audit gives nhs.uk a score of 87, with nothing critical, a single warning and a small handful of notices across more than a hundred passed checks. The technical section is a clean sweep — every check passed. Almost everything the site gives back is in speed; on-page loses a single check and mobile loses nothing at all.
Read that as a description rather than a grade. This is a site with no clever tricks in it. No rendering gymnastics, no compressed sitemaps, no interstitials between a URL and its content. The templates are dull and identical and they load. When you have 51,900 near-identical pages, dullness is the feature: whatever is true of one page is true of all of them, and there is no room for a per-page mistake to hide.
It also puts nhs.uk squarely in the company of the other well-built sites in this series — a shade above GOV.UK, a shade below MDN and itch.io. Nothing exceptional, nothing broken. The exceptional part of this site is not its engineering.
Copy this in an afternoon
You almost certainly do not have a national health service behind you. You can still borrow the structure.
1. Count your own composition. Fetch your sitemap index, fetch every child, and group the URLs by their first path segment. Not a sample — all of them. Most teams discover their site is mostly one thing they never talk about, and that thing is usually where their links already are.
2. Ask what your readers need next. If your explanatory content ends by telling someone to go and do something, build the pages that let them do it. The handoff pages are usually cheaper to produce than the explanations and they attract a completely different kind of link.
3. Check that your directory is still true. Take a random sample from every large sitemap file — not the first entries, a scattered sample — and check the status codes. Do it per file, because decay is rarely evenly spread. One category rotting while the others are fine is the normal failure mode.
4. Separate entries from pages. Deduplicate your sitemap URLs before you quote a page count anywhere. If one file exists only to attach markup to pages listed elsewhere, it adds zero pages, and counting it inflates every ratio you build on top.
What you cannot copy
The unfair advantage is monopoly of the underlying record. nhs.uk does not have to compete for the canonical list of GP surgeries — it is where that register is published. No commercial directory can assemble the same data with the same authority, and no competitor can offer a better version of a fact that only one organisation is entitled to state. The link rate follows from that, not from clever content design. A site earning 5 referring domains per page at this scale is not being rewarded for its writing. It is being cited because it is the register.
And here is what we could not check:
Our tool returned no domain rank for nhs.uk. The Backlink Analyzer came back with an empty rank field, exactly as it did for GOV.UK earlier in this series. That is a defect in our tool, not a statement about the site, and we are publishing no rank rather than guessing one. Every other figure in this article from the same run reconciles internally — the dofollow and nofollow counts sum exactly to the reported total — so we have no reason to distrust the rest of the report.
The pharmacy decay figure is a sample, not a count. We checked 150 pharmacy URLs, not 11,220. The proportion that fails is an estimate with real uncertainty around it, and the live-pharmacy figure derived from it is an estimate too. We are confident the effect is real and pharmacy-specific, because the other three profile sets showed nothing like it across their own samples. We are not confident to the nearest hundred.
Page counts and sitemap entries are not the same thing. A small number of URLs are listed in two sitemap files at once, which makes the raw entry total slightly higher than the number of distinct pages. We have used the raw total where we are describing the sitemap and flagged the distinct total in the FAQ. The gap is under half a percent and changes no conclusion, but a reader adding up columns deserves to know it is there.
We have no traffic, revenue or ranking data. Everything here comes from public sitemaps, public pages and our own tools. We do not know how many people use the directory versus the encyclopedia, which sections bring in search traffic, or how the organisation weighs the two internally. The composition of a sitemap tells you what a site publishes. It does not tell you what anybody reads.
We only looked at one country's site. The pattern of dead pharmacy listings hints at providers who sit under a different administration within the health service. We did not measure any other health-service site, so nothing here should be read as a claim about them.
The NHS pharmacy listings are a reminder that a sitemap is a claim, and claims go stale quietly. Run the UNmiss Website Audit on your own site and see what your crawler is being sent to fetch.
- Crawls your pages and flags broken and redirecting URLs
- Separates critical issues from warnings and notices
- Scores on-page, technical, speed and mobile independently
Frequently asked questions
How many pages does nhs.uk actually have?
The sitemap tree contains 56,889 entries, of which 56,698 are distinct URLs. The difference is 191 URLs that appear in two files at once: all 176 entries in the video sitemap are re-listings of pages already in the health-information file, plus 15 overlaps between the search-landing file and the same. Use 56,889 when you mean sitemap entries and 56,698 when you mean pages.
What is the full breakdown by file?
Service directory profiles 25,787; pharmacy profiles 11,220; GP surgery profiles 8,489; dentist profiles 6,404; health information 4,037; campaigns 730; video-bearing pages 176; find-a-service landing pages 46. The four profile files together are 51,900 entries, or 91.2% of the total.
Was this a full count or a sample?
A full count. The index at the site root has 8 children, none of which is itself an index, so all 8 leaves were fetched and every entry counted. Each file's URL blocks matched its location tags exactly and each ended with a well-formed closing tag, so nothing was truncated. The three largest files were re-fetched and returned identical counts, and an independent second pass reproduced every figure exactly.
How was the content grouped by type?
By verifying the path prefix of every URL, not by sampling. The four profile files are completely pure — all 51,900 of their URLs sit under exactly one path prefix each, with no exceptions. The health-information breakdown by topic is likewise an exact count of all 4,037 URLs by first path segment.
What are the largest health-information topics?
Medicines with 1,412 pages and conditions with 1,241 lead by a wide margin. After those come tests and treatments at 317, mental health at 163, NHS services at 146 and live well at 135, with the remaining topics tailing off from there across 30 distinct segments in total.
How bad is the pharmacy listing decay?
Of 150 sampled pharmacy URLs, about 12% returned a not-found response and a further 4% or so redirected to an online-only pharmacy path, which is a live page rather than a failure. That puts the estimated live pharmacy count around 9,900 against 11,220 listed. The same sampling method found no failures at all across 220 service-directory, GP and dentist pages.
What did the backlink analysis return?
297,601 referring domains and 20,745,693 total backlinks, split 17,493,455 dofollow and 3,252,238 nofollow — 84.3% dofollow. The dofollow and nofollow figures sum exactly to the reported total, so the report is internally consistent. Our tool returned no domain rank for this site, which is a defect on our side.
How does the audit score compare to other sites here?
nhs.uk scored 87, with 0 critical issues, 1 warning, 3 notices and 117 checks passed. Technical was perfect at 67 out of 67, on-page 24 out of 25, mobile 2 out of 2, and speed 25 out of 29. For comparison in this series: PyPI 91, MDN 88, itch.io 88, GOV.UK 86, Eventbrite 85, levels.fyi 83, Meetup 83, Rotten Tomatoes 80 and Healthline 74.
The lesson here is not that you should build a directory. It is that most sites already contain one and have never thought of it that way. A catalogue of products, a list of venues, a register of members, a set of location pages nobody has touched in two years — these are the pages that get cited by people who have no interest in your writing, and they are usually the pages nobody on the team is responsible for.
nhs.uk has 13 of those pages for every article it publishes, and it earns links at a rate that almost nothing of its size matches. The encyclopedia gets the reputation. The phone book does the work.
The catch, as the pharmacy listings show, is that a directory has to stay true or it quietly stops being one. If your own sitemap is making promises about the world, the cheapest thing you can do this week is find out how many of them still hold — start with the UNmiss Website Audit.
Measured on 29 August 2026. Page composition comes from a full census of all 8 nhs.uk sitemap files, independently recounted; the pharmacy availability figures come from a 150-URL random sample and are estimates.
Backlink and audit figures come from the UNmiss Backlink Analyzer and UNmiss Website Audit, run on the same date. Our backlink tool returned no domain rank for this domain, so none is published.