Stop 410-ing Old Pages: The URL Decay Decision Tree for AI-Era B2B SaaS SEO

The reflex, and what it costs you now

A page underperforms. Traffic slid, rankings faded, nobody links to it internally anymore. The reflex is to reach for a 410 and be done with it. Clean up the index, tidy the site, move on.

That reflex quietly forfeits four things at once.

You lose the backlinks pointing at that URL. You lose the index status Google already granted it. You lose the internal links feeding it authority from the rest of your site. And in the AI era you lose one more asset that did not exist a few years ago: the AI-citation slot that answer engines already expect to find at that address.

Here is why that last one bites. Google has a longstanding practice of continuing to crawl 404 URLs "just in case those pages were removed by accident and have been restored." A dead URL is not forgotten. It keeps getting checked, and that repeated recrawl is Google's system holding the door open for you to put content back. A 410 tells Google the opposite: the removal is intentional and permanent, so it stops checking as often.

Now layer the AI engines on top. Ahrefs analyzed 16 million unique URLs cited by ChatGPT, Perplexity, Copilot, Gemini, Claude, and Mistral. AI assistants land users on 404 pages 2.87x more often than Google does. ChatGPT is the worst offender, with 1.01% of its clicked URLs and 2.38% of all the URLs it cites returning a 404, against Google baselines of 0.15% clicked and 0.84% cited. The models do this because they lean on training-cutoff data instead of a live fetch, so they surface pages that were moved or deleted.

Read those two facts together. AI engines are already over-serving URLs that no longer exist. When you 410 an indexed page, you are not cleaning up a dead end. You are creating one at an address a model may already be trying to send people to.

The real question is not "which status code"

Most of the 404-versus-410 debate is a distraction, because neither code is a penalty. Google states plainly that it does not penalize websites for 404 status codes. Its own crawling-errors documentation lists 404 and 410 side by side as equivalent signals that "the page doesn't exist and you don't want search engines to index the page." John Mueller's summary is that the processing difference between the two is so minimal he cannot think of a time he would prefer one over the other for SEO.

So stop arguing about the code. The status code is the last decision you make, and the least consequential one. The real question is upstream: should this URL die at all?

Decayed traffic is a symptom, not a verdict. Content decay is the gradual loss of a page's rankings as it goes stale or competitors pull ahead, and the usual fix for decay is refreshing the page, not deleting it. A page can be underperforming today and still be sitting on backlinks, an index slot, and latent AI demand that a rewrite would reactivate. Killing it forecloses all of that to save yourself a couple of days of index cleanup.

The URL Decay Decision Tree

When a page decays, run it through this tree in order. Stop at the first branch that fits. The default is to keep the URL alive; a true kill is the exception you have to earn.

Ask these three questions, in order:

1. Does the topic still serve your ICP? If yes, keep the URL and rewrite it in place. That is Branch A, the default. If no, keep going.
2. Is there a stronger sibling page that already owns this intent? If yes, merge the decayed URL into it. That is Branch B. If no, keep going.
3. Is the page off-strategy, with no relevant merge target and an unrescuable slug? If all three are true, the page qualifies for a true kill. That is Branch C. If not, default back to Branch A and rewrite the angle.

Branch A. Rewrite in place (the default)

The topic still serves your audience, so the URL stays. Rewrite the content at the same address: refresh the data, fix the angle, rebuild it around a liftable answer. You keep every backlink, the index history, the internal links, and the AI-citation slot, and you spend that equity on a better page instead of forfeiting it.

Expect a temporary dip in Google Search after a major rewrite. That is normal. Re-crawl and re-evaluation of a heavily changed page play out over several weeks. In our own work we typically see a rewritten URL recover over roughly four to eight weeks, and requesting reindexing can shorten the wait. If you want the rewritten page to earn citations, structure it for extraction. See our guide to structuring content for AI extraction and how long AI citations actually take.

Branch B. Merge into a stronger sibling

The topic is off-strategy on its own, but a stronger page on your site already owns the same intent. Redirect the decayed URL to that specific sibling with a 301. A permanent redirect is a canonicalization signal to Google, and 3xx redirects do not lose PageRank. Gary Illyes confirmed back in 2016 that any 301, 302, or 3xx redirect passes its full link value, which retired the old belief that each hop cost roughly 15% of authority. The backlinks and equity flow into the target instead of evaporating.

The guardrail: the target has to be genuinely relevant. Redirect a removed URL to an unrelated page or to the homepage and Google treats it as a soft 404. It recognizes there is no topical relationship, effectively treats the original as a 404, and drops it anyway, redirect or not. A 301 preserves value only when the destination actually covers the same intent. If nothing on your site does, do not force a merge.

Branch C. The true-kill, last resort only

Reserve the 410 for the narrow case where all three conditions hold at once:

1. The topic is genuinely off-strategy, with no place in your roadmap.
2. No topically relevant page exists to absorb a 301.
3. The slug is unrescuable, so a rewrite would be a stretch on the URL itself.

Only then does removal make sense, and even then a plain 404 does nearly the same job. A true kill is where you deliberately accept the loss of the backlinks, the index slot, and the AI-citation candidacy, because nothing is left worth preserving. That is a deliberate call on a handful of URLs, never a batch default.

The three options, side by side

Anti-patterns that quietly cost you rankings

Do not blanket-301 dead pages to the homepage. It feels tidy and it fails. Google reads a redirect to an irrelevant page or the homepage as a soft 404 and drops the original URL, so you get none of the equity transfer you were after. Redirect to a specific, relevant page, or leave the URL to 404 and rewrite it later.

Do not mass-410 or mass-prune. Content pruning is a relatively niche tool, best suited to very big sites or sites carrying a lot of irrelevant, low-quality content, and its benefit is mostly crawl-budget efficiency that is not guaranteed. If you do prune, do it in batches and monitor the impact before any sweeping change, and never delete a page that holds backlinks or traffic. Redirect those to a close match instead. When we faced a batch of low-quality, AI-generated posts on our own site, we rewrote them in place rather than mass-killing the URLs, because bulk deletion would have thrown away index slots and inbound equity for pages that were fixable.

The AI-era corollary: you are manufacturing your own dead ends

There is an inbound version of this problem and an outbound version, and they are the same mechanism pointed in opposite directions.

Inbound: when an AI bot fetches a URL on your site that 404s, the model has pattern-matched your URL structure and inferred that a page should exist there. That is latent content demand. Build the page it expected and you can earn a citation within weeks. We wrote up how to capture that signal in the AI demand engine, a free Cloudflare-to-Notion pipeline that logs which URLs AI bots hit and miss.

Outbound is the flip side. AI engines already send users to dead URLs far more often than Google, because they surface pages from training data that were moved or deleted. When you 410 an indexed page, you delete a destination those models may already be trying to reach. You are not tidying the web. You are producing the exact dead end AI engines over-produce on their own, and you are doing it to a URL that had standing.

This is the reasoned position behind the whole tree: a 410 does not just remove a page from Google today. In our analysis it permanently forfeits that URL's latent AI-citation value, because the URL an engine expected to find is now gone for good, and Google stops recrawling it to see if it came back. That inference is ours, drawn from how 410, recrawling, and AI link behavior interact. It is not a line from Google. It is the logic of the evidence, and it is why the reflex 410 is a worse trade than it looks.

Decayed does not mean dead. Rewrite it, merge it, and only kill it when there is truly nothing left to save.