What is a 410 error?

The 410 belongs to the 4xx family, the errors attributed to the client's request, and shares the basic message with the 404: there is nothing left to serve at that address. The difference lies in the certainty. The 404 does not say whether the absence is passing or final, while the 410 asserts that the condition is probably permanent. The specification hands out the roles without ambiguity: if the server does not know, or has no way to determine, whether the absence will last, the code that applies is the 404.

It is worth reading who that message is written for. The specification describes the 410 as an aid to web maintenance, meant to signal that the resource was withdrawn on purpose and that whoever links to it from outside would do well to remove that link. The natural recipient is therefore the rest of the internet, not the search engine. And the mark is neither mandatory nor lifelong: the specification leaves it to the person responsible for the server to decide how long to keep it, or whether it is worth setting at all.

Google's documentation on status codes grants the 410 no special treatment. It says the opposite, with a sentence that settles the debate: all 4xx errors except the 429 are treated the same. The crawler informs the next system that the content does not exist, indexing removes the URL if it was indexed, addresses first discovered already carrying that error are never processed, and the crawl frequency of that path drops off little by little.

From this follows the important part, which is what does not happen. There is no shortcut. To learn that a URL answers 410, Google has to request it again, so the pace is set by how often it crawls that particular address and not by the number of the code. A URL it visits every few days will disappear sooner than one it visits every few weeks, and that holds exactly the same way for a 404. Google publishes no deadline either, for one or for the other.

HTTP/1.1 410 Gone
Content-Type: text/html; charset=UTF-8

When the urgency is real, say personal data published by mistake or a wrong price spreading around, the fast route is a different one. Search Console includes a removals tool that acts within about a day, with a limitation worth understanding: the removal lasts around six months and is a temporary hiding, not a deletion. It buys time while the real change, the one on the server, propagates on its own.

If the effect on the index is practically the same, it is fair to ask why bother. The answer sits on the side of whoever runs the site.

A 404 report mixes things that have nothing to do with each other: typos in hand typed addresses, internal links broken by a template change, paths that fell over during a migration, and pages somebody decided to delete. The first three are faults that need fixing. The fourth is a decision already taken. Marking that fourth category with 410 separates it from the noise, and from then on a spike of 404 in the server logs means something concrete again.

The second reason is the one the specification gives: telling whoever links from outside that the resource was withdrawn on purpose. It is a courtesy with a practical effect, because a 404 invites a retry and the assumption of a passing failure, while a 410 closes the matter.

What should not be expected is a ranking effect. Applying 410 in bulk does not appreciably speed up any index cleanup and improves nothing on its own. If the problem is a bloated index, the lever lies in deciding which pages are surplus and why, not in the number they return.

• Reserve the 410 for deletions already decided, where there will be no replacement and the address will not be reused later.
• If an equivalent page exists, a 301 redirect remains better than any deletion code, because it puts the links to use instead of throwing them away.
• Before applying 410 in bulk, measure how many external links point at those URLs. The ones still bringing real visits deserve a destination.
• If the page must stay reachable for people but out of the index, the noindex tag is the right tool, not an error code.
• Check that URLs marked with 410 are not blocked in robots.txt, because an address that cannot be crawled cannot deliver its status code either.
• Faced with real urgency, combine the change on the server with the Search Console removals tool, and note the date on which that removal expires.

• Expecting the 410 to work as a delete button. Google's documentation grants it no speed advantage whatsoever over the 404.
• Applying it wholesale to sold out product pages that will return to the catalogue next season, instead of keeping the URL alive with a note on its status.
• Marking as deleted addresses that still hold valuable external links, rather than redirecting them to the nearest equivalent page.
• Using 410 to pull pages out of the index that users still need, when what was called for was a noindex.
• Blocking in robots.txt the very URLs that have just been given a 410, so that Google never gets to see it.