This article was AI-generated as part of an experimental historical-content project. The date reflects the period being analyzed rather than the date the article was originally written.
Two pieces published today describe the same quiet labor. The Washington Post reports that a warning, “This article may incorporate text from a large language model,” has been placed on hundreds of Wikipedia articles over the past year, on subjects from Danish nationalism to Natalie Portman. The Verge describes a new speedy deletion rule English Wikipedia’s volunteers have adopted for articles that are clearly machine-written and were never reviewed by the person who submitted them.
Most corporate readers will file this under Wikipedia housekeeping. I think it is a case study in where the real defense of a company’s public description sits, and it is not where most companies assume.
What the editors changed
Normally, an article nominated for deletion goes through a seven-day discussion. The new rule lets administrators skip that step when an article is clearly AI-generated and unreviewed. According to The Verge, they look for three signs: text addressed to the user (“Here is your Wikipedia article on…”), nonsensical citations with wrong authors or publications, and references that do not exist at all, such as dead links, ISBNs with invalid checksums or DOIs that resolve to nothing.
Separately, WikiProject AI Cleanup keeps a list of phrases and formatting habits typical of chatbot prose, while cautioning that those habits alone should not decide a deletion. One reviewer quoted by The Verge said editors are “flooded non-stop with horrendous drafts.” Another said the “lies and fake references” take “an incredible amount of experienced editor time to clean up.”
The Wikimedia Foundation, which does not write the community’s rules, is building tools around them. Marshall Miller, a product director there, compared the community’s response to an “immune system.” The Foundation is developing Edit Check, which reminds new contributors to add citations and keep a neutral tone, and a planned Paste Check that asks people who paste in a large block of text whether they wrote it.
Why this is an entity story
The failure modes the editors are hunting are exactly the ones that hurt companies. A fabricated citation attached to a claim about a business looks, to a casual reader, like a real one. A fluent paragraph about a company’s history can mix accurate facts with an invented acquisition date or a lawsuit that belongs to someone else. Once that sentence sits in an article, it does not stay there. Wikipedia text is reused widely, by search features, data aggregators and the models that answer questions about the company.
Early last year I wrote that Wikipedia had not agreed on a policy for LLMs and asked who would audit the sentences AI repeats. Part of the answer, it turns out, is volunteers checking ISBN checksums by hand.
The firewall has a capacity
The lesson for companies is not that Wikipedia has this under control. It is that the barrier between synthetic prose and the neutral sentence that seeds other systems is made of people, and people have finite hours. The faster junk arrives, the longer a bad sentence may survive before someone with the time and the expertise catches it. Articles about mid-sized companies and their executives, which have fewer watchers than articles about celebrities, may wait longest.
What follows for companies
Treat a Wikipedia article as something to reread periodically, not a project that ended when it looked right. Check the history for large recent additions and for the LLM warning banner. Confirm that every citation about your company resolves to a real source that says what the article claims.
And do not add to the pile. A communications team or agency that drafts Wikipedia text with a chatbot and pastes it in without checking every reference is producing exactly what these volunteers are deleting. The right path has not changed: disclose the affiliation, propose changes on the talk page, bring independent sources.
The new deletion rule is a small procedural change. What it reveals is that the most important quality check on a company’s public description is often a volunteer deciding, one case at a time, whether a citation is real.