Docs · AI Detector
AI Detector.
Paste any text and see the mechanical tells that make it read as machine-generated, named and highlighted. No black box, no percentage.
How it reads
Most AI detectors return a confidence percentage from a model you cannot inspect, and they are wrong often enough to get real people falsely accused. This one does the opposite. It runs the deterministic engine behind Typographer, matching the concrete habits of machine writing, and shows you every one it finds.
The verdict is qualitative, not a grade:
- Reads human — no tells, with natural sentence rhythm.
- Mostly human — a few tells to clean up, or a flat rhythm with no machine vocabulary.
- Reads machine-written — the mechanical habits below are doing the talking.
What it flags
- Empty AI vocabulary: delve, robust, seamless, leverage, tapestry.
- Throat-clearing, signposting, significance inflation, chatbot artifacts, sycophancy.
- The flat sentence rhythm that machine writing falls into.
Read it honestly. This is deterministic tell-detection, not a probability guess. A clean result is not proof a human wrote it, and a flagged result is not proof of cheating, because people use these phrases too. Do not use it to accuse anyone.
Runs entirely in your browser. Nothing is uploaded.