A creator finds an old post that is wrong now.
It may be embarrassing.
It may be genuinely unsafe, because a claim has quietly aged past the point where anyone checked it.
The quick answer is to delete the page.
That removes it from a menu.
It doesn’t decide what should happen when someone still arrives at the old link.
A public page is more than its current wording.
It has an address, a sitemap entry, inbound links, a browser cache, an edge cache, and a place in search.
Take the text away without choosing a status and every one of those systems is left guessing.
An archive is not a spare room for pages nobody wants to explain.
So start with the plainest possible question.
What should a reader find when the old link still arrives?
If something genuinely took over the job, redirect it.
Make the redirect direct — one hop, to the page that actually replaces the old one.
A URL shouldn’t wander through a chain of successors, and a crawler will give up on a long one.
A tidy redirect to an unrelated current page isn’t care.
It is a broken answer wearing a neat status code.
If nothing replaces it, say so.
Gone is a real answer. It tells readers and crawlers the resource is permanently retired, which is more honest than an empty page still reporting success.
It also leaves room for the record underneath to survive while public rendering stops.
Then take the retired page out of navigation, out of internal search, and out of the sitemap.
Those are discovery decisions.
They are not proof the record has ceased to exist, and they shouldn’t be confused with one.
Some failures are too deep for a redirect.
A fundamental error, borrowed text, uncleared material: those call for a retraction, and a retraction is a specific shape.
Keep the address.
Mark the title.
Put the reason at the top, where nobody can miss it.
Preserve an unchangeable copy of the original.
Then tell search engines not to index it while leaving the links through it intact.
A retraction isn’t a reward for an error.
It is a readable record of one.
Whatever you decide, the change has to reach the copy people actually receive.
A page can sit in an edge cache for a day and in a browser for an hour after you have changed it.
Switch a page to a redirect, a gone response, or a retraction without a purge and the old version stays public for exactly as long as the cache says it can.
So issue the purge with the status change, not after someone complains.
The page is not retired until the cached page is gone too.
Preservation begins where public access stops.
A stored file is not safe because it is stored. Bits rot, drives fail, and a silent corruption looks identical to a healthy file until you open it.
So fingerprint the copy when you save it, and check the fingerprint again on a schedule.
A mismatch means the stored bits changed, and it means you still have time to repair from another copy.
Keep more than one copy, and not all in one place.
For a captured web page, keep something that preserves the page as a page — the markup, the styling, the media — rather than a screenshot of it.
A screenshot proves what it looked like.
It doesn’t let anyone follow what it said.
Formats age, too.
A file that opens today may need a different format in a decade, so plan the migration before it becomes urgent, and keep the original alongside the migrated copy.
The migrated file is for use.
The original is evidence of what existed.
And deletion is a status like any other, not a faster archive.
Some records can’t be destroyed, and some must be. Know which one you are holding before the delete becomes irreversible.
So choose the status before removing the page.
Redirect what genuinely has a successor.
Retire the dead end honestly.
Retract what needs to stay visible as a warning.
Preserve what someone may still have to prove.
A record can leave the menu without leaving history.