How the diagnostic works, where it breaks, and why we tell you both.
The diagnostic doesn’t render a verdict. The method pages below explain how the profile is built, what each pattern means, where the signals fail, and how we calibrate against real writing.
A profile with many flags across many families means stronger overlap with AI-typical surface patterns. It is not proof of authorship, cheating, or intent, and it is never enough to skip human review. For the full interpretation framework, see How It Works. If policy is involved, read Limitations first.
Evidence before enforcement
Don’t turn the diagnostic into a decision without checking the evidence. Do AI Detectors Work? covers the reliability case. The False Positive Hall of Fame shows that false positives happen to real writing by real people. The research publishes the data behind every detection rule.