Wikipedia discussion: “Signs of AI writing”

The current Wikipedia page, “Signs of AI writing,” is an advice resource—not an official policy—created by the WikiProject AI Cleanup. It catalogs common writing and formatting patterns observed in AI-generated text, especially writing produced by large language models (LLMs) like ChatGPT, to help Wikipedia editors detect undisclosed AI-generated content.

Key points:

• The list is descriptive, outlining patterns seen in AI-written text, not prescriptive rules for Wikipedia writing.

• It emphasizes that these signs alone do not prove content is AI-generated, as LLMs are trained on human writing, and human editors may share some of these habits.

• Examples of typical AI-generated patterns include:

  • Overuse of generic, positive, or exaggerated descriptors

  • Repetitive summary/conclusion phrases (e.g., “In summary,” “Overall”)

  • Rigid formulaic structure, especially in sections like “Challenges” or “Future Prospects”

  • Excessive use of boldface, lists, or title case in headings

  • Superficial or vague analyses and attributions (e.g., “some critics argue”)

  • Phrasal templates left unedited (like fill-in-the-blank instructions)

  • Use of curly quotes/apostrophes, emojis in headers or lists, and negative parallelisms (e.g., “not only… but also…”)

  • Knowledge-cutoff disclaimers and prompt refusals appearing in text (“As an AI language model…”)

The page also cautions against relying solely on AI-detection tools, highlighting the importance of human judgment, and connects these surface issues to deeper policy and sourcing risks for Wikipedia if left unaddressed.