Wikipedia editors have published one of the most detailed public guides on how to identify AI-generated writing, offering a clearer framework at a time when readers increasingly suspect that what they’re seeing online was produced by large language models.
The document, titled “Signs of AI writing,” is part of Project AI Cleanup, an effort launched in 2023 to help editors manage the surge of AI-written submissions appearing across the platform. With millions of edits flowing into Wikipedia daily, volunteers have had a vast sample size to study, allowing them to identify patterns that automated detection tools struggle to catch.
Wikipedia flags linguistic habits that AI tools consistently fall into
The guide notes that algorithmic detectors are “basically useless,” and instead focuses on recurring stylistic traits that appear frequently in AI-generated text but rarely in neutral, encyclopedic writing.
One consistent pattern is an emphasis on why a subject is important, often framed in generic or inflated terms. Phrases such as “a pivotal moment,” “a broader movement,” or descriptions that attempt to elevate minor events appear far more often in AI submissions than in human-written Wikipedia entries.
The guide also highlights a specific grammatical pattern: clauses that describe an action as “emphasizing,” “underscoring,” or “reflecting” some vague importance. Editors say this “present participle” structure shows up repeatedly in AI-authored drafts and becomes easy to spot once you know to look for it.
Vague marketing tone is another red flag
Editors point out that AI models often default to language common across marketing copy, travel blogs, and product descriptions — all of which are abundant in training data. As a result, landscapes become “scenic,” views are “breathtaking,” and settings are described as “clean” or “modern,” regardless of context.
“It sounds more like the transcript of a TV commercial,” the guide notes.
This tendency to embellish contrasts with Wikipedia’s strict tone guidelines, which prioritise neutrality and verifiable facts over lyrical description or subjective value judgments.
A guide that could influence how readers judge online writing
The document has gained wider attention after writer Jameson Fitzpatrick highlighted it on X. It suggests that some linguistic tendencies may persist across AI models, even as they grow more advanced and less reliant on obvious filler phrases.
The broader implication, Wikipedia editors argue, is that AI writing — no matter how sophisticated — carries foundational patterns shaped by the data it learns from. Those traits can be masked, but not entirely eliminated.
As public awareness of those signals grows, it could reshape how newsrooms, educators, and platforms assess the authenticity of text online.
Featured image credits: Wikimedia Commons
For more stories like it, click the +Follow button at the top of this page to follow us.