Twenty-five patterns, ranked by how much they give you away.
A language model writes whatever is most likely to come next, so by default it makes the choice that fits the widest range of readers. A person chooses for one reader. Every pattern below is one form of that default. The first five justify an edit on a single sighting; the weaker ones only count when several appear together, because a careful writer might use any one on purpose.
Drawn from Wikipedia’s Signs of AI writing, maintained by WikiProject AI Cleanup.
Staging instead of stating
These are the strongest and most frequent tells in current model prose. Act on one sighting.
- 01Not X but YThe negative half names something no one claimed, so the positive half sounds larger.
- 02One-line closers and dramatic fragmentsThe line asks the reader to pause on a claim instead of adding to it.
- 03Sayings that sound deepAn ordinary point is dressed as a hidden truth or an aphorism, and the dressing adds no detail.
- 04Staged run-up before the pointThe writer announces the point or stages a moment of candor instead of making the point.
- 05Arguing with no oneThe text answers an objection or rejects an option that appears nowhere else, usually a leftover from an earlier draft.
Rhythm by rule
A person may do any one of these on purpose, so the weaker ones need company from other tells.
- 06Forced triadsIdeas arrive in threes to sound complete, whether the meaning has three parts or not.
- 07Repeated sentence openingsSeveral sentences in a row start with the same subject, often she or he, because repetition is handled by rule instead of by ear.
- 08Dashes as the universal connectorweak aloneA dash lets the writer skip choosing how two clauses relate, so a model reaches for it everywhere.
- 09Stacked qualifiersRepeated editing adds one qualifier after another until every claim sounds uncertain, usually to repair an earlier overstatement rather than to report real d…
- 10Hyphenated pairs everywhereThese pairs are hyphenated in every position.
- 11Passive voice and missing subjectsThe text hides who acts or drops the subject.
Inflation and borrowed authority
The fact underneath is usually sound. Keep it and remove the dressing.
- 12Overused AI wordsModels use these words far more often than people do, especially in groups.
- 13Inflated significanceAn ordinary detail is said to mark a change, prove a legacy, or promise a future.
- 14Vague connection or associationThe text says two things are connected without saying how.
- 15Shallow -ing ridersAn -ing phrase is bolted onto a simple fact to make it sound deeper.
- 16Sales languageThe text reads like an advertisement, especially for places, culture, products, or organizations.
- 17Borrowed authorityA name or an unnamed authority stands in for what was said.
- 18Avoiding is, are, and hasSimple verbs are replaced with longer phrases.
Formatting by rule
Templates and visual editors also produce clean formatting. The tell is decoration on every item.
- 19Bold as decorationWords are bolded without a reason, and vertical lists give every item a bold label and a colon.
- 20Decorative headingsHeadings capitalize every main word, and headings or list items carry emojis or arrows (→) as decoration.
- 21Curly quotation marksweak aloneCurly quotes (“...”) appear where the writer or target format uses straight quotes ("...").
Leftovers from the chat and the draft
Remove these outright. Nothing here needs rewriting.
- 22Chatbot residueA chatbot's greeting, praise, offer, or closing remains in text that should stand on its own.
- 23Knowledge-limit disclaimers and guessesThe text mentions where the model's knowledge ends, or admits it found no source and then fills the gap with a plausible guess.
- 24A heading repeated in the first sentenceA heading is followed by a one-line paragraph that restates it before the real content begins.
- 25Writing about the previous versionDocumentation and comments describe what the text replaced instead of the current behavior.
See which of these are in your draft.
Paste it in. Every match gets marked, named, and explained.
Open the editor