Content pruning is the systematic review of a site's pages to decide, one by one, whether to leave them as they are, rewrite them, merge them into another page or remove them from the index. The gardening image only goes so far: a dead branch is visible to the naked eye, a dead page is not.
We tried to count them. Of 4,646 pages from six UK websites that currently appear in Google's top 100 for at least one search, 2,660 bring in less than one estimated organic visit a month: 57.3%. They are the natural candidates for pruning. For almost none of them is deleting the right answer.
Google explains why in its own documentation.
Google says deleting is the last thing to do
On the Search Central page about core updates, under what to do when organic traffic drops, there is a sentence that leaves no room for doubt: "Deleting content is a last resort, and only to be considered if you think the content can't be salvaged". The same document warns that removing entire sections of a site is usually a sign that those sections were written for search engines rather than for people.
In August 2023 CNET deleted thousands of old articles. An internal memo obtained by Gizmodo explained that removing content sends Google a signal that the site is fresh and relevant. Danny Sullivan, Google's Search Liaison, replied publicly on 8 August: "Are you deleting content from your site because you somehow believe Google doesn't like 'old' content? That's not a thing!". The next day he clarified the useful part: cleaning up is always fine, but you should not assume that deleting something just because it is old will magically improve the site's SEO. John Mueller closed the argument from another angle: a page getting few visits does not mean it is not useful to the people who read it. The three statements are collected and dated by Search Engine Roundtable.
This is the difference that makes content pruning useful or harmful. Removing pages is a consequence of the decision, not the decision. The decision is working out what job each URL is doing.
How many dead pages a real website has
Most of the figures quoted on pruning come from US sites with millions of URLs. To get an order of magnitude for ordinary sites we measured six real UK websites of very different sizes: two marketing publishers, a marketing training site, an SEO software company, an engineering group and a mid-sized engineering manufacturer.
For each one we counted how many pages appear in the top 100 on Google UK for at least one keyword, how many of these bring in less than one estimated organic visit a month and how many bring in at least ten. Data from the UK database of DataForSEO Labs, 26 September 2026.
| Site | Pages in top 100 | Under 1 visit/month | % | At least 10 visits/month | % |
|---|---|---|---|---|---|
| marketingweek.com | 3,272 | 1,869 | 57.1% | 360 | 11.0% |
| econsultancy.com | 574 | 348 | 60.6% | 57 | 9.9% |
| smartinsights.com | 553 | 328 | 59.3% | 97 | 17.5% |
| renishaw.com | 169 | 73 | 43.2% | 56 | 33.1% |
| sitebulb.com | 64 | 35 | 54.7% | 9 | 14.1% |
| hepcomotion.com | 14 | 7 | 50.0% | 6 | 42.9% |
| Total | 4,646 | 2,660 | 57.3% | 585 | 12.6% |
Three things stand out from this table. First: the share of near-invisible pages is similar on sites that differ in size by more than two hundred times, between 43% and 61%. Second: only one visible page in eight passes ten estimated visits a month, so most of the traffic always sits on a minority of URLs. Third, and most important for most readers: a typical company site has dozens or a few hundred pages to assess, not hundreds of thousands. One of the pages with the most estimated traffic on Renishaw's site in our extraction is a page about Isambard Kingdom Brunel, which ranks for seventeen searches, on the site of a company that makes metrology equipment.
Scale matters because it changes what the job is. With thirty pages to review, you read every URL. With four thousand, you work by samples and rules, and the wrong rules do damage in proportion.
Crawl budget, the reason almost none of you can use
Many of the guides that rank for content pruning list crawl budget among the reasons to do it: fewer useless pages, more of Googlebot's time spent on the good ones. Search Engine Land's guide promises better crawlability in its own description. In the abstract it is true. Google, however, has written down in black and white who it applies to.
The official guide to managing crawl budget is aimed at sites with more than a million pages that change once a week, or sites with more than ten thousand pages that change every day. Google says these are not exact thresholds, but that is the order of magnitude. Semrush, in its pruning guide, lowers the bar to ten thousand pages.
Renishaw, the Gloucestershire engineering group, has 169 pages visible on Google UK in our extraction. HepcoMotion has 14. Neither has a crawl budget problem, and nor does the typical small or mid-sized business website. If someone proposes a mass prune to free up crawl budget, ask how many pages your site has before saying yes.
This does not make pruning pointless. It changes its purpose: it is not a technical fix for crawl efficiency, it is an editorial decision about what the site claims to know. It is the same reasoning behind an SEO audit when you look at the whole rather than the single page.
Where it has really worked, and by how much
Documented cases with numbers do exist, and they mostly come from very large sites. We report them with their source, because the proportions do not transfer to a fifty-page site.
In the collection of cases published by Ahrefs, Francesco Baldini describes a vehicle valuation platform cut from 4,860,000 to 1,500 pages, with organic visits up 160% and conversions up 105% within weeks. Eugene Zatiychuk of Belkins removed 400 pages, about two thirds of the site, between January and March 2023, working one folder at a time: organic traffic went from 3,000 to almost 10,000 monthly visits. Victor Pan, then at HubSpot, removed 3,000 pages from the sitemap and saw indexing times drop from hours and days to minutes. Bryan Casey removed more than a thousand pages from IBM's navigation: the goal was usability rather than traffic, and navigation NPS rose by 30%.
Four cases, four orders of magnitude far removed from an average company website. What transfers is not the growth percentage, it is the method: all four worked in batches rather than in one go, and three of the four had a problem with duplication or mass-produced content, not with old content.
In 2026 zero clicks does not mean zero value
The unwritten rule of every pruning guide is: if a page brings no clicks, it is not needed. Until two years ago that held. Today Search Console measures a shrinking part of what happens to a page.
Ahrefs analysed 863,000 searches and 4 million URLs cited in AI Overviews. In January 2026 only 37.9% of citations came from pages in the top ten organic results; in July 2025 the share was 76%. 31.2% came from pages in positions 11-100 and 31.0% from URLs that do not even appear in the top 100. The study is published by Ahrefs; the sample is international and not segmented by country, so the figure should be read as a trend rather than a measurement of any single market.
The practical consequence is direct. A page that does not appear in any useful position can be cited as a source by an AI assistant, and that citation does not show up as a click in Search Console. Before removing a page because it brings nothing, check whether a generative engine is using it, with the tools we cover in the guide to AI Overviews ranking factors and in the piece on the impact of AI Overviews on SEO. It is a check that did not exist in 2023 and that now changes the decision on one page in three.
The page we would have deleted by mistake
In 2020 the website of Macropix, an LED screen manufacturer based in Milan and a Visilay client, ranked in position 88 on Google Italy for "monitor pubblicitario" (advertising display), a search with 390 queries a month. Position 88 means page nine of Google: zero clicks, zero useful impressions, zero reasons to keep it under any automatic pruning rule.
That page is now in second position. Not because of a technical fix: it was rewritten by the people who make those displays. On the same keyword cluster, in July 2026 macropix.it held a 25.55% share of voice, ahead of Amazon at 18.08%. The full story is in our article on SEO for manufacturing.
A page in position 88 and a useless page have exactly the same numbers in Search Console. Only one question tells them apart: does the company have something real to say on that topic? If yes, the page should be rewritten. If not, it should go. No numerical threshold answers that for you.
Four actions, in this order
The order matters. Keeping costs nothing, updating costs little, consolidating costs a redirect, and removing is irreversible in practice even when it is reversible in theory. This is the grid we use, with the last twelve months of Search Console data in front of us.
| What the last 12 months of data say | Action | Why |
|---|---|---|
| Clicks falling, impressions stable or rising | Update | The demand is still there, the page no longer captures it |
| Zero clicks, average position between 11 and 30 on relevant queries | Update | It is one page away, not dead |
| Zero clicks, zero impressions, topic already covered better elsewhere on the site | Consolidate | Two pages on the same question weaken each other |
| Two or more URLs alternating on the same query, none in the top 10 | Consolidate | Cannibalisation: one page wins more than two halves |
| Zero clicks, zero impressions, no inbound links, topic outside the commercial scope | Remove | There is nothing to recover |
| Content that has expired by nature: past event, closed promotion, old price list | Remove or archive | Keeping it online is only worth it if someone still searches for it |
| Zero clicks but cited by an AI assistant, or linked from external sites | Keep | The value is there, Search Console does not measure it |
Two warnings about the data. Impressions should be read over a full twelve months, otherwise seasonality skews the judgement: a page on "corporate Christmas gifts" looked at in July seems dead, and in the UK that search goes from 110 queries in July to 1,900 in November (DataForSEO, Google UK, September 2026). And the number of backlinks should be checked before, not after: removing a page that receives links from outside throws away authority that someone else chose to give you.
What to return when you remove a page
A stubborn myth circulates here: a 410 supposedly removes the page from the index faster than a 404. Google's documentation on status codes says the opposite: "All 4xx errors, except 429, are treated the same".
| Situation | What to return | What Google does |
|---|---|---|
| There is a page that answers the same question | 301 to that page | Follows the redirect and uses it as a strong signal for the destination |
| There is no equivalent on the site | 404 or 410, either one | Removes the URL from the index if it was indexed |
| The page is useful to users but should not be in the SERP | noindex | Excludes it, but keeps requesting it: you save no crawl budget |
In the same crawl budget guide Google adds that a 404 is a strong signal not to recrawl that URL, while a blocked page stays in the queue much longer. If you have decided a page should no longer exist, make it respond as if it does not. We have two dedicated guides on how a 301 redirect works and on handling broken URLs.
The mistakes that turn pruning into damage
- Redirecting everything to the home page. Google treats a redirect to a non-equivalent page as a soft 404. The result is the same as a 404, plus a messy line in the Search Console report.
- Deleting in bulk and measuring afterwards. If you remove 200 pages on the same day and traffic drops, you will never know which ones caused it. Belkins went one folder a week for exactly this reason.
- Trusting clicks without looking at impressions. A page with zero clicks and three thousand impressions has a title and search intent problem, not an existence problem.
- Forgetting internal links. Every removed page leaves broken links on the pages that cited it. Fix them in the same round of work, not the next one: this is the part of the job that makes an internal links review worthwhile.
- Expecting results in two weeks. Google writes that some changes show up within a few days, but that it can take several months for its systems to register that the site as a whole has improved.
How often to do it again
Semrush recommends batches every one to three months for large sites and one or two reviews a year for small ones. On the projects we run, a different rhythm works better: one full review a year, and a quarterly check limited to pages that have lost clicks while keeping their impressions. That group is small, takes an hour to review and almost always contains the pages worth rewriting rather than removing, as the section on updating explains in the guide on creating SEO content.
If the site has fewer than a hundred visible pages, content pruning as a standalone project does not exist: it is normal editorial maintenance, done alongside the review of the company blog. Above a thousand pages it becomes structured work, with a spreadsheet, an owner and a review date. If you need someone to do it with you, it is part of what we do in our SEO projects.
What nobody counts when they delete
There is a cost of pruning that no guide mentions: the pages you remove are also the ones the site learnt something from. In long projects the article that brings the most leads today is often the third rewrite of a piece that did not work in its first two versions. If the first version had been deleted at the annual review because it had zero clicks, the third would never have existed. Before removing a page, ask whether it is dead or just on its first attempt.
Frequently asked questions
