Wikipedia publishes its content under a license it cannot take back. Every article carries CC BY-SA 4.0, and older text still carries the GFDL beneath it — both let anyone copy, republish, and reuse the material, indefinitely, without asking permission first.
That permission is what built the ecosystem of sites that run on Wikipedia without being Wikipedia: mirrors that reproduce whole articles, aggregators that pull summaries into search results, apps that ship a frozen copy of the encyclopedia for offline reading, and tools trained on a dump of the database from whatever month someone downloaded it. None of them are obligated to keep their copy current, and most do not try.
A correction on the live article does not reach any of them. A biography can carry an inaccurate birth year for three years, get it fixed in an afternoon once a source finally surfaces, and still turn up with the old year on a mirror that copied the page in 2023 and never ran its import script again. The fix is real. It is also local — to one article, on one site, as of one date.
Re-users are not required to keep their copies up to date, and most reusers do not; a republished copy reflects the article as it stood on the day it was copied, not as it stands now.
There is no deletion discussion to open on a mirror, no talk-page template to post, often no listed owner to contact at all. Wikipedia's own machinery — revision history, talk pages, the noticeboards that handle a disputed claim — has no jurisdiction past the edge of the site that built it. A downstream copy answers to whoever built it, on whatever schedule they chose, if they are still maintaining it at all.
The practical result is two separate clocks. One runs from the moment an edit lands on the live article — fast, often the same day a source is added or a claim is removed. The other runs from whenever each downstream copy last refreshed, which can be months, a product cycle, or never. Treating the first clock as the whole job is the mistake we see most often: a client is told the article was fixed, checks a search result a week later, finds the old wording still sitting in a snippet pulled from a mirror, and reasonably wonders whether anything happened at all.
Working the second clock looks less like editing and more like tracking: knowing which mirrors and aggregators actually rank for a subject's name, which ones still sync against Wikipedia and which stopped years ago, and which snippets will age out on their own without anyone's intervention. None of it reverses what made the original correction necessary, and none of it is unique to unflattering material — a flattering paragraph propagates through the same channels, on the same uneven schedule, for exactly the same reason.
