Wikipedia editors have published a public guide titled "Signs of AI writing," detailing characteristics observed in AI-generated text submitted to the platform. This initiative stems from "Project AI Cleanup," an ongoing effort launched in 2023 to address submissions from large language models.
The guide, which Wikipedia editors developed based on millions of daily edits, departs from the use of automated detection tools, describing them as largely ineffective. Instead, it focuses on specific linguistic patterns and stylistic habits that are common in AI output but rare in typical human-authored Wikipedia contributions.
Key indicators highlighted by the guide include a tendency for AI-generated text to emphasize a subject's importance in generic terms, using phrases like "a pivotal moment" or "a broader movement." It also notes the inclusion of minor media mentions to artificially inflate a subject's perceived notability, a characteristic more often found in personal biographies than independent sourcing.
Further stylistic quirks identified involve "tailing clauses" that attach hazy claims of significance, such as statements asserting an event is "emphasizing the significance" or "reflecting the continued relevance" of an idea. Additionally, the guide points to the pervasive use of vague marketing language, where descriptions like "scenic landscapes" and "breathtaking views" are common, leading editors to liken the prose to "the transcript of a TV commercial."
Since its inception in 2023, Project AI Cleanup has engaged Wikipedia's extensive editor community in systematically identifying and managing AI-generated submissions. The resulting public guide suggests that these observed AI writing habits are deeply embedded in the training and operational frameworks of current AI models.
