How it works

You paste your text. It goes to a fast AI classifier along with our own detection criteria, which we built from published research on how AI writing differs from human writing. You get back a probability that the text was written by AI, usually in under a second.

Think of it as a quick gut check. Got a stack of papers and no time to read them all closely? Run them through, and if one or two come back strongly AI, that’s the one worth a closer look. Every result lands somewhere on this scale:

• 🧑 Appears human — reads like a person wrote it.
• ✍️ Likely human — mostly human patterns, with a few polished spots.
• 🤷 Too close to call — mixed patterns, could go either way. We say so instead of guessing.
• 🤝 Possibly AI‑assisted — some AI patterns, like human writing edited or rewritten by AI.
• 🤖 Appears AI — strong AI patterns throughout.

Clean, polished writing is not counted against you. That assumption is what gets good writers falsely flagged by other checkers.

We keep the details of what we look for private. Publishing them would just hand people a checklist for getting AI text past the detector.

Longer text gives a more reliable answer, which is why we need at least 700 characters and let you check up to 20,000.