Free tool · no sign-up
Free AI style detector
Paste a text and see its estimated AI-style score (0–100): the density of typical AI clichés in this page's language, sentence rhythm, vocabulary diversity and repetition. The analysis runs in your browser — the text is never sent anywhere.
Honestly: this is a stylometric estimate, not proof of authorship. No AI detector is fully reliable — see the research.
How we calculate this score — and what we don't promise
The score is a stylometric estimate: a deterministic sum of explicit text features — including the density of typical AI phrases in this page's language (“it's important to note”, “in today's fast-paced world”), uniformity of sentence rhythm, lexical diversity and repetition (plus a few minor style signals). It's the same formula the humai humanizer shows as “Est. AI style” before and after rewriting. The formula always assumes a margin of uncertainty — in practice scores fall roughly between 20 and 96, so even the most human text will never get a zero.
What we don't promise: this score is not proof that a text was (or wasn't) written by AI. No detector — ours included — offers that certainty; research shows false accusations even against texts written by humans. Details and sources: AI detector accuracy — what the research says.
Frequently asked questions
Can this detector prove that a text was written by AI?
No — and no tool can. Our score is a stylometric estimate: a transparent sum of explicit text features (including AI clichés typical of this page's language, monotonous sentence rhythm, low vocabulary diversity, repetition). A high score means the text SOUNDS machine-like — not that it was definitely produced with ChatGPT.
How is the 0–100 score calculated?
With a deterministic heuristic — the same formula the humai humanizer uses for its “before” and “after” scores. It analyses, among other things, the density of AI phrases typical of this page's language, uniformity of sentence length, lexical diversity (MATTR) and trigram repetition. The same text always gets the same score; the formula assumes a margin of uncertainty, so in practice scores fall roughly between 20 and 96.
Is my text sent anywhere?
No. The analysis runs entirely in your browser — the text is not sent to any server or any AI model, and we never store it.
Are AI detectors (GPTZero, Turnitin, JSA) reliable?
Research shows that every detector makes mistakes — including on texts written by humans, especially in languages other than English. Treat scores (ours too) as a probability signal, not a verdict. We collected the data and sources in our article on the accuracy of AI detectors.