See what reads as AI, then fix it
Paste a draft and get the specific words, sentence patterns and formatting habits that make writing sound machine-generated. Not a probability score: the actual phrases, and what to do about them.
What it checks
- 722 over-indexed terms, graded into four reliability tiers
- 27 sentence patterns: negative parallelism, tricolon autopilot, pseudo-profound closers
- 23 formatting tells: pasted assistant preambles, markdown leakage, hidden Unicode
- Sentence rhythm, measured as burstiness against the human range
- Formulaic openers and engagement-bait sign-offs
- Spelling, grammar and readability in the same pass
Detectors are built for the reader. This is built for the writer.
Originality.ai, GPTZero and Turnitin sell to the person checking somebody else’s work: a teacher, an editor, a hiring manager. What they want is a verdict about a document they did not write, and a percentage is a reasonable answer to that question.
If you used AI to draft something, you already know the answer. A score tells you nothing you did not know and gives you nothing to act on. The useful question is not "is this AI" but "which parts read that way, and what should they say instead".
So this shows you the phrases, underlines them in place, and rewrites them on request. And it is honest about the limits: no tool can promise you a number on somebody else’s classifier, and anyone claiming otherwise is guessing. Removing the habits detectors look for usually moves those scores. The point is that it reads like a person wrote it.
Frequently Asked Questions
One pass, before you publish
Spelling, grammar, readability and AI phrasing in a single check, and the rewrite that fixes them. No second subscription, and nothing you write is ever sent to a language model.
Open the studio: 5 free rewrites a day