AI Text Detector
Paste text. See every signal we used to decide.
Most AI detectors give you a percentage and a vibe. This one shows you the actual measurements - burstiness, vocabulary diversity, AI-cliché density, transition-word patterns - and tells you what each one means. No invented certainty.
About this detector
AI detection is not a solved problem. We treat it that way.
Every paid detector on the market - GPTZero, Originality, Copyleaks - has a 5–20% false positive rate on human text. Stanford research has shown that detectors flag writing by non-native English speakers as AI roughly half the time. OpenAI shut down their own detector for being too inaccurate to ship.
Our approach is different: we never tell you “87% AI”. We show you three bands - likely human, mixed, likely AI - and the raw signal values that put the text in that band. You see the math, not a black box.
Use this as one input among many. A creative writing teacher with thirty years of experience is still better than any detector. So is reading the text yourself.