Content Pruning in SEO: Why I Deleted Only 5 of 201 URLs

My site lost about 99% of its Google traffic across three core updates. Pruning is the only thing that has moved it back.

So I have done this work properly, on my own domain, with money and years on the line. And here is the part that surprised me.

I went through 201 old URLs one at a time. I deleted five.

Every guide on page one of Google tells you to find your weak content and cut it. Google's own documentation says the opposite, in plain words: "Deleting content is a last resort, and only to be considered if you think the content can't be salvaged."

Not one of the six pages ranking above this one quotes that line. This post is the decision I actually ran instead.

What is content pruning in SEO?

Content pruning is going through your site page by page and deciding what to do with the weak ones: improve, merge, redirect, keep but isolate, or remove.

It is not "delete your thin content". Deletion is one outcome of five, and it should be the rarest.

The reason to bother is mechanical. Google assesses quality partly at the site level, not only page by page. That means a pile of weak pages can hold down pages you never touched.

This is why pruning can lift a page you did not edit. You did not improve that page. You stopped dragging it down.

Weak posts are only one source of this. Filters and sort parameters generate the same problem at a much larger scale, which is why faceted navigation is worth auditing before you start deleting articles.

That site-level effect is also why the timing matters. The September 2023 Helpful Content Update used a sitewide classifier, March 2024 moved helpfulness to page-level assessment inside core ranking, and the June 2025 core update pushed it back towards sitewide judgement again. Pruning bites harder now than it did in 2024.

How do you find pages worth pruning?

Most guides tell you to sort Search Console by clicks and cut the bottom. That is how you delete pages that were fine.

Use four numbers together, over a full twelve months, not one. The most defensible published set comes from Seer Interactive's pruning case study, where removing roughly 14,000 low-value pages helped turn a multi-year decline positive. Their floor was:

A page has to fail on all four to be a candidate. Failing one is noise.

Then two guardrails. Never prune a page under six months old, because new pages need three to six months before their numbers mean anything. And never treat low traffic as proof of low quality.

John Mueller drew that line himself when asked about well-written but unpopular news articles: "So from that point of view, I wouldn't necessarily call those articles low quality articles." Judge the page, then the numbers.

The pattern that actually matters: ranking but not wanted

Here is a real slice of my own Search Console, 28 days to 19 July 2026.

Look at the bottom of that chart. kfc-franchise-cost sits at position 12.1 with 5,004 impressions and a 0.16% click-through rate. google-adsense-approval ranks 9.9 with 2,225 impressions and 0.18%.

Those pages are not failing to rank. They rank fine. Nobody wants them from me.

That is a different diagnosis from a dead page, and it needs a different fix. A dead page gets removed. A ranking-but-unwanted page is either an intent mismatch you can fix in the title, or a page that has no business being on an SEO consultancy's site at all.

Which action does each page need?

Five outcomes, not two. This is the decision I run on every page.

Backlinks first, then demand, then relevance. In that order, because the order protects the things that are hard to get back.

Improve

The topic still has real search demand and the page is simply bad. This is where most salvageable pages land, and it is the outcome Google's guidance pushes you towards.

It means a genuine rewrite with first-hand checks, not a refreshed date. My directory submission sites page was a dump of URLs. Same intent, rebuilt with real verification.

Consolidate

Several pages compete for one intent. Merge the best parts into the strongest URL and 301 the rest into it.

A quick test: if two pages share seven or more of the same URLs in their top 10 results, they are the same page as far as Google is concerned. Merge them.

Redirect

The page has backlinks but the content is irrelevant to what you do now. 301 it to the nearest genuinely relevant page so the link equity survives.

The trap here is the one most people fall into. An irrelevant 301 is treated as a soft 404, and passes nothing. If there is no relevant destination, do not invent one.

Fence

This is the outcome almost nobody writes about, and it saved me a lot of traffic.

Some pages earn real visits but say nothing about your brand. On my site those are franchise guides and money-online roundups. Deleting them throws away traffic; keeping them in the main site blurs what the brand is about.

So I fence them. They live under /resources, out of the navigation, out of the entity graph, on generic Article schema. They keep earning without telling Google that an SEO consultancy is partly about KFC franchises.

Remove

No backlinks, no impressions, no relevance, no way to salvage it. Now, and only now, delete it.

When should you use 410 instead of 301?

Not one of the six pages ranking for this term mentions 410. Search Engine Land's guide says to "always use a 301 for deleted content", which is wrong for genuinely dead pages.

Here is the actual hierarchy.

410 means "Gone". It is a stronger, clearer signal than 404, and Google drops those URLs faster. That is the whole point when you are trying to clear low-quality pages out of an index that is being judged as a whole.

Whichever you pick, set it at the server and keep it to a single hop. If you are doing this on WordPress, I have written up the redirect methods that survive a theme change separately.

Of my 201 legacy URLs, exactly five got a 410. They were pages like bypass-facebook-verification-process-2018, which is dead on the topic and dead on the year.

What does pruning do to your AI visibility?

This is the part every framework on this topic misses, and it changes the decision.

To appear as a supporting link in AI Overviews or AI Mode, Google states a page "must be indexed and eligible to be shown in Google Search with a snippet". There is no separate index for AI features.

So noindex does not just remove a page from blue links. It removes it from Google's AI answers too. If you reach for noindex as a soft alternative to deleting, understand that you are switching off AI visibility as well.

The second problem is worse, because it is invisible. Every framework on page one tells you to prune on Search Console clicks. Search Console cannot tell you whether ChatGPT or Perplexity already cite that page.

A page with almost no Google clicks may be quietly feeding AI answers. Pruning on clicks alone is now a partially blind instrument, and nobody selling you a pruning process says so.

I do not have a clean fix for this yet. What I do is check the obvious candidates by hand in ChatGPT and Perplexity before I remove anything with real impressions, and I 301 rather than 410 when I am unsure. That is a workaround, not a solution.

If a page is earning citations rather than clicks, that is an argument for rewriting it, not removing it. The signals that make a page worth citing in ChatGPT and Perplexity are not the same ones that win a click in Search.

What actually happened when I pruned my own site?

196 URLs took a 301. Five took a 410. Of the content I kept, 91 posts were rewritten into /insights and 72 were fenced into /resources.

That is 2.5% deleted. Every guide on this SERP is oriented around the 2.5%.

The results, honestly. The site bottomed at 285 clicks in February 2026, then ran 684, 612, 1,197 and 1,573 through June. On a good day it now does 80 to 100 clicks.

That is a real climb. It is also still roughly 96% below where the site once stood, so I am not calling it a recovery.

And I have to be straight about attribution: the rebuild is not even live yet. This movement happened on the old WordPress site. What changed was pruning, cutting and thinning the weakest pages. So this is Google reassessing a less-bad site, not a rebuilt one.

A partial climb on thin content is fragile. The next core update could take it back. I am writing the number down anyway, because a pruning guide with no numbers in it is exactly the problem with this SERP.

How do you prune without breaking things?

The process matters as much as the decision. This is a technical SEO job, so treat it like one.

Then wait properly. Google's guidance is blunt about the clock: some changes show in days, but "it could take several months for our systems to learn and confirm that the site as a whole is now producing helpful, reliable, people-first content".

Do not judge a prune in a fortnight. Judge it against the next core update.

Common mistakes I see

Deleting because a page is unpopular. Unpopular is not low quality. A good page serving a small audience is not a candidate.

Redirecting everything somewhere. An irrelevant 301 is a soft 404. It passes nothing and wastes the link.

Pruning right after a core update. You are reading noise. Wait until the update finishes, then wait another week.

Treating a date change as an improvement. Changing updatedAt without changing the content is detectable, and it is the exact "quick fix" Google's documentation warns against.

Pruning once. Content decays continuously. This is a job for every six to twelve months, not a project.

Final take

Content pruning works. On my site it is the only thing that has worked so far, which is why I take it seriously enough to argue with how it is usually taught.

But the framing on page one of Google is backwards. Pruning is not a deletion exercise with a few exceptions. It is a quality decision with five outcomes, and deletion is the rarest and least recoverable of them.

Start with backlinks, then demand, then relevance. Improve what has demand, merge what competes, fence what earns but does not fit, and delete only what is genuinely dead.

If you are pruning to hit a deletion target, you have already got it wrong. I ran 201 URLs through this and deleted five, and I would still call it aggressive.