The AI writing tells
Nearly every list of AI writing tells on the internet draws on one source: Wikipedia's Signs of AI writing, maintained by its AI Cleanup project. It is the largest catalogue of its kind and it is genuinely good.
Almost nobody quotes the four things it says about itself, and they change how you should use it. It is descriptive rather than prescriptive. It is an advice page, not policy. Its sourcing is uneven term by term. And as of August 2026 it carries a banner saying its own coverage of recent models is out of date.
This page sits under brand voice for AI. If you would rather write your own prohibitions than learn somebody else's list, eleven questions turn them into a file a checker can run.
What does the biggest catalogue actually say about itself?
Four statements, all on the page, all easy to check, and all routinely dropped by the pages that cite it.
- It is descriptive, not prescriptive. In its words: it consists of observations, not rules.
- The signs are not the problem. It says in bold that the patterns are only potential signs of a problem rather than the problem itself, and asks readers not to treat the signs as the things to fix.
- It expects false positives by design. It notes that not all text with these markers is machine-written, because the models were trained on human writing, including Wikipedia itself.
- It is scoped to Wikipedia. It describes itself as a field guide for spotting undisclosed AI content on Wikipedia, and says only some of the signs are broadly applicable.
- It is an advice page of WikiProject AI Cleanup and not Wikipedia policy, because it has not been reviewed by the community for consensus.
None of that makes it less useful. It makes it a different tool from the one most people quote it as: a description of what machine writing has tended to look like, not a test you can run on a colleague's draft.
Why does the list keep expiring?
Because the tells are model artefacts, and models change. The catalogue dates its own vocabulary by model era, and the shape of that is the most useful thing on it.
| Era | Words listed | Examples |
|---|---|---|
| 2023 to mid-2024 | 18 | delve, intricate, interplay, garner, bolstered |
| Mid-2024 to mid-2025 | 12 | the era of large language models moving on |
| Mid-2025 onwards | 4 | emphasizing, enhance, highlighting, showcasing |
Eighteen down to four. The catalogue also records the fate of the most famous one: delve was heavily overused in 2023 and early 2024, became less frequent later in 2024, then dropped off sharply in 2025.
So a list of AI words has a half-life, and most of the lists in circulation are quietly reporting the vocabulary of a model nobody uses any more. The page itself agrees: as of August 2026 it carries a maintenance banner asking for its coverage of the most recent models to be updated.
That is worth sitting with before running any of these words as a checker. A rule that flags delve in 2026 is testing for a model that has been superseded twice.
Which of these are actually evidenced?
Unevenly, and the page is honest about it if you look. Delve carries four separate citations to published work. Other entries carry none: key, as an adjective, is tagged with a citation-needed marker.
That is the difference between a term somebody measured and a term somebody noticed. Both can be true. Only one is evidence, and a list that presents them in the same typeface is inviting you to treat them as equally solid.
If you take anything from this page into your own style rules, take the cited ones and leave the rest as observations.
Which of these can a checker catch, and which cannot?
A word list is trivially checkable. A construction usually is not, and the difference decides what can go in a style file at all.
| Kind of tell | Checkable? | Why |
|---|---|---|
| A specific word or phrase | Yes | String matching, and it is what a substitution rule is for |
| Sentence length or rhythm | Yes | A metric rule counts it |
| Punctuation habits | Yes | Existence rules catch the character |
| Hedging and overqualification | Partly | The common phrasings are listable, the habit is not |
| Empty symmetry, the not-this-but-that shape | Partly | Some forms match, most do not |
| Confident vagueness | No | Requires knowing whether the claim is true |
| Being generic | No | Requires knowing what specific would have been |
The bottom two rows are where most of the actual problem lives, and no linter is going to help with either. That is not a reason to skip the top rows. It is a reason not to believe a green run means the writing is good.
The rules that do survive into a document are the mechanical ones, and the download includes a style checker configured with them.
Why is a tell not proof?
Because the base rate defeats it. These constructions appear in human writing constantly, they appeared there first, and the models learned them from us. A marker that is common in both populations cannot separate them, however strongly it feels like a signal.
The catalogue says as much, which is why the honest use of it is to improve your own writing rather than to judge somebody else's. This site will not claim a file makes writing undetectable, and it will not sell you the reverse either.
What should you actually do about it?
Write the prohibitions down and let a tool enforce them, rather than trying to recognise a vibe after the fact. A banned-words list you chose is worth more than a tells list somebody else observed, because it encodes your judgement instead of a model's habits.
Eleven questions produce that list, with a checker already configured, and nothing you type is transmitted while you do it.
Where these come from
Every claim above is quoted from one of these, and each was read on the date beside it. If one of them has changed since, the page is wrong and we would like to know.
- Wikipedia: Signs of AI writing, and its own caveats checked 2026-08-17
- Wikipedia: WikiProject AI Cleanup checked 2026-08-17
The tells expire. Your own rules do not: write yours down once and hand them to every assistant.