Citations don't expire on Wikipedia, but the pages they point to do. A newspaper redesigns its site, a company retires an old microsite, a government portal renumbers its addresses, and the link behind a years-old footnote starts returning an error. The fact it supported hasn't changed. The claim was true and checkable the day someone cited it; only the address failed.
The project has a name for this, link rot, and a standing instruction for handling it: a dead link is a broken address, not proof the source never existed. Bots, InternetArchiveBot chief among them, patrol English Wikipedia's citations on a schedule, check whether a URL still resolves, and when it doesn't, search the Wayback Machine for a snapshot taken near the citation's own date. Where one exists, the bot adds an archive link to the citation template and tags the original as dead rather than deleting anything. Where none exists, the citation is flagged for a human to look at, and the sentence beside it becomes a target.
We watched that second case play out on a client's page last year: a profile piece from a regional business weekly, cited twice, went dark when the publication folded its archive into a paywalled relaunch. Nobody noticed for the better part of a year. Then an editor doing routine cleanup on a stub-adjacent article flagged both citations as dead, and three weeks later an unrelated deletion discussion cited those same two dead links as evidence the subject's coverage was thinner than the article implied. The underlying story had never changed. What had changed was whether anyone reading the discussion could still check it in two clicks.
A dead link doesn't erase what it once supported; find an archived copy before treating the source as gone.
The practical habit sits earlier than any of this. When we file a citation into a draft or an edit request, we pull an archive.org snapshot the same day and add it to the template alongside the live link, rather than waiting for a bot to do it after the link has already failed and the sentence has already been challenged. Archiving a weak source doesn't strengthen it; a press release captured on day one is still a press release. But a genuinely independent, in-depth profile we fail to archive is one accident, a site migration, a bankruptcy, a change of ownership, away from becoming unverifiable for a reason that has nothing to do with the fact it once supported.
