How does GPTZero detect AI paraphrasing and AI bypassers?

GPTZero can detect AI-written text that was rewritten to disguise that origin. That includes text run through an AI bypasser or humanizer, and text a person reworded by hand.

Scan the document with Advanced Scan. When the model classifies it as AI paraphrased, the result reads: “We are highly confident this text was originally AI, but rewritten by AI or a human.” A badge above that line says Possible AI Paraphrasing. The result card still marks this as a beta feature.

At lower confidence, the opening changes to “moderately confident” or “uncertain.” Open the AI score to see the split between chance AI Generated and chance AI Paraphrased.

The model detects AI-paraphrased text from certain popular AI tools with a false positive rate of less than 0.1%. Coverage of bypasser tools is still improving.

Detection is not infallible. For a record of how a document was written, use GPTZero Authorship, including the Google Docs Writing Report in the GPTZero browser extension.