Tuesday, October 21, 2025

Indicators of AI writing on Wikipedia – FlowingData


From WikiProject AI Cleanup, a information on recognizing faux writing on Wikipedia.

This checklist isn’t a ban on sure phrases, phrases, or punctuation. Nobody is taking your em-dashes away or claiming that solely AI makes use of them. Not all textual content that includes the next indicators is AI-generated, as the big language fashions that energy AI chatbots are skilled on human writing, together with the writing of Wikipedia editors. That is merely a catalog of quite common patterns noticed over many 1000’s of situations of AI-generated textual content, particular to Wikipedia. Whereas a few of its recommendation could also be broadly relevant, some indicators—notably these involving punctuation and formatting—might not apply in a non-Wikipedia context.

Extra on em-dashes:

Whereas human editors and writers typically do use em dashes (—), LLM output tends to make use of them extra typically than nonprofessional human-written textual content of the identical style, and makes use of them in locations the place people are extra doubtless to make use of commas, parentheses, colons, or (misused) hyphens (-). LLMs particularly have a tendency to make use of em dashes in a formulaic, pat method, typically mimicking “punched up” sales-like writing by over-emphasizing clauses or parallelisms. LLMs overuse em dashes as a result of they have been skilled (typically illegally) on novels, and novelists have all the time used em dashes extra typically than is typical of a layperson.

This signal is most helpful when taken together with different indicators, not by itself.

I feel I’ve been subconsciously utilizing extra commas as of late.

Related Articles

Latest Articles