Treating an automated score as proof that writing is AI-generated leads to unfair calls.
AI Detector gives an English-only statistical estimate with a signal breakdown, framed as indicative rather than a verdict.
How to use AI Detector
- Paste at least a 40-word English passage into the box.
- Keep the signal breakdown on to see the patterns behind the score.
- Read the score as one indicator and review the passage yourself.
Use cases
- Sanity-checking your own draft before submitting it.
- Understanding which patterns make text read as machine-written.
- Comparing two passages to see which reads more uniform.
Good to know
AI Detector looks at statistical patterns common in AI-written English, such as uniform sentence length, cliche phrasing, repetition and low word variety, and turns them into a 0 to 100 estimate. It is not proof: human writing can score high and edited AI text can score low, so treat the number as a prompt to look closer. It needs about 40 or more words, only supports English, and analyzes the first 40,000 characters of very long text with a visible note.
Frequently asked questions
Can a high score prove text was written by AI?
No. The score is a statistical guess from surface patterns; confident human writing can score high, so it is never proof on its own.
Why does it need at least about 40 words?
Short passages do not give enough sentence-length and word-variety signal for a meaningful estimate, so very short text is unreliable.
Does it work for languages other than English?
No. It is an English-only heuristic; non-English text is detected and skipped rather than scored.
What does the signal breakdown show?
It shows the four patterns behind the score, so you can see whether uniformity, cliche phrasing, repetition or low word variety drove the result.