Site search on UK public sector websites put the right page first in 65% of the 3,287 searches we could score in October 2026. In 18% of those searches, the right page was not in the top five results at all.
By "the right page" we mean the page the website itself links to for that task from its own menu or sitemap. We typed short terms a visitor would use, such as "jobs", "contact us" and "term dates", into the search box on each homepage and recorded where that page appeared. No organisation is named, and any web team can repeat the method.
How often does public sector site search find the right page?
Public sector site search found the page a site links to within the top five results in 2,702 of 3,287 scored searches (82%), across 832 sites. It came first in 2,125 searches (65%) and second to fifth in 577 (18%). The remaining 585 searches (18%, about one in six) missed the top five.
Misses were spread widely. Of the 832 sites we could score, 334 (40%) had at least one of their test terms miss the top five. One miss on one of up to five terms was enough to count, and a flagged miss is not always a clear failure.
How we ran the test, and what it cannot tell you
A browser loaded each homepage, found the site's own search box and typed five short terms chosen for that sector. A term was scored only where the site's own menu or sitemap proved the right page existed.
We scored 832 of 1,168 sites. The other 336 are excluded from every figure here. On 13 we found no site search. Another 48 were unreachable or blocked automated access. On the remaining 275 a search box was found, but no term could be scored, because the results page could not be read reliably, the search errored, or the site's own menu and sitemap did not prove which page was right.
The test is an automated screen. Our scraper flagged each miss automatically. When we checked a sample of flagged misses by eye, about one in three was a clear failure, so the 18% miss rate is an upper bound on clear failures. It does not judge whether the other results were useful, and we have no data on how often real visitors type these terms.
Which search terms missed the right page?
"Jobs" was one of the hardest terms we tested. Across 265 searches, the jobs page the site links to was missing from the top five 31% of the time and came first in only 45%. Searches for "contact us" missed the top five in 26% of 181 searches.
Both "jobs" and "contact us" were used across several sectors, which makes them the fairest terms to test on your own site. Single-service terms did far better. "Bins" put the right page first in 87% of 262 searches, and "council tax" in 83% of 252. "Complaints", the term with the most scored searches, came first in 69% of 303 and missed the top five in 13%.
The weakest results sat in university and college terms. For "term dates" and "open days" the right page was in the top five only about half the time (56 of 119 and 60 of 120). Some of these may be course-finder boxes that only search courses, which a visitor can mistake for site search just as our scraper did. A homepage box that looks like site search but only searches one thing is a findability problem in its own right.
Sector results reflect the terms each sector was tested on
Each sector was tested on terms suited to it, so the comparison below mixes the sector with its terms. Council sites were tested on single-service terms such as bins, council tax, pothole and parking permit.
| Sector | Scored sites of sites in scope | Scored searches | Right page first | Missed top five |
|---|---|---|---|---|
| Councils | 266 of 380 | 1,111 | 81% | 3% |
| NHS trusts and bodies | 239 of 275 | 930 | 59% | 17% |
| Central government bodies | 102 of 133 | 388 | 61% | 20% |
| Universities and colleges | 164 of 244 | 607 | 51% | 39% |
| Fire, transport and port bodies | 51 of 114 | 219 | 52% | 30% |
| National museums | 10 of 22 | 32 | 50% | 16% |
Two rows need care. Fire, transport and port bodies had only 51 of 114 sites scored. National museums had 32 scored searches on 10 sites, too few to compare with anything.
A search that returns nothing is the plainest failure
In 93 of 3,287 scored searches (about 3%), the site returned no results at all for a page it links to from its own menu. Of those 93, 66 were on university and college sites, 11% of that sector's 607 searches, and the course-finder caveat applies directly here.
A zero-result page suggests to a visitor that the information does not exist. On "term dates" it happened 13 times in 119 searches, and on "accommodation" 18 times in 129.
Why site search misses the right page
The test recorded where the right page appeared. The causes below are possibilities to check on your own site.
A jobs page may sit on an external recruitment portal that the site's search does not index. Contact details may be spread across several pages, none titled with the words a visitor types. Older news stories, PDFs and committee papers can mention a term more times than the service page does. GOV.UK's own publishing guidance tells writers to "use popular terms in the titles, headings and first paragraphs of your content" to make it easier to find. A page that never uses the visitor's word gives a website search engine less to match on.
Configuration matters as well. W3C's technique G161 describes a search function that can spell-check terms, include different word endings and allow for synonyms. On Drupal, the Drupal Search API module handles word endings through a stemmer for English, one of its processors; others include a stopwords list and type-specific boosting, which "Adds a boost to indexed items based on their datasource and/or bundle". Boosting service pages can lift them above news. Our piece on the hidden cost of decentralised content publishing covers how duplicate pages build up.
Axistwelve built Axistwelve Semantic Search for the case where visitors type their own words and the page they need uses different ones. The product page describes how it approaches that gap.
How can you test your own site search in an afternoon?
You can test your own site search in an afternoon with ten terms and a spreadsheet. The steps are derived from our method, widened to ten terms.
- Take five terms from your search analytics and five from your top tasks or main menu. Prefer the short words a resident types, such as "jobs" or "contact us", over your page titles.
- For each term, write down the right page before you search: the one your menu, sitemap or A to Z links to. If you cannot name one page, note that as a finding.
- Open the homepage logged out, in a private window, on desktop and on a phone. Check whether the first search box searches the whole site or only courses, jobs or documents.
- Type each term as a visitor would and record where the right page appears: first, second to fifth, lower, or absent. Record zero-result pages separately.
- Where the right page did not come first, note what did.
- Repeat three terms with a misspelling or plural to test spelling suggestions and stemming.
If you use Google Analytics 4, its enhanced measurement records a search when a results page URL carries a query parameter such as q, s or search. Check whether automated visits reach those reports before trusting the top terms; our piece on what bot traffic costs you covers the wider cost.
What to ask a site search supplier
Ask a supplier how it proves relevance, and expect a test you can repeat. When GOV.UK moved to a new search engine, its team used judgement lists, "sets of ratings for how relevant results are for a given search query". That is the professional version of the test above.
| Question | A strong answer | A weak answer |
|---|---|---|
| How will you test relevance before and after go-live? | A list of known queries with agreed right pages, run before launch and after each change, with results shared | The engine handles relevance automatically |
| What can our content team change without a developer? | Promoted results, synonyms and boosts by content type in an admin screen, with a change log | Every change is a support ticket |
| What does the search index? | A stated list, including PDFs and external portals such as jobs, with the re-index frequency | Everything |
| Does the homepage box search the whole site? | Yes, or a clear label and a route to whole-site search | A course or document finder labelled "Search" |
| How do we leave? | Synonyms, promoted results and configuration exported in an open format | Configuration lives only in the vendor's console |
On Drupal, ask which module or service runs search, which processors are switched on, and who maintains it through upgrades.
Before you change your site search
Run the ten-term test first and keep the results as your baseline. Fix the cheap misses next: retitle a page with the visitor's word, add a synonym, or demote the old item that outranked it. Check the zero-result and top-search reports in your search analytics every month, and re-run the test after each platform change. If a wider rebuild is coming, our piece on what a future-ready council website platform looks like covers the platform decision.
Frequently asked questions
Does WCAG require a public sector website to have site search?
WCAG does not require site search on its own. Success Criterion 2.4.5 Multiple Ways, at level AA, requires more than one way to locate a page within a set of pages, and a search function is one sufficient technique.
Which search technology do public sector sites use?
On 646 of the 832 sites we scored (78%), site search ran on the search built into the content management system or on a server-side setup we could not attribute to a named vendor. The remaining 186 used other detected setups, mostly named search products or plugins. Our data does not link any technology to results.
How much is site search used on public sector websites?
We have no usage data of our own. GOV.UK reported in October 2024 that its site search is used more than 4 million times a month. Your own search analytics will show what visitors type.
