Signals are evidence, not a verdict
AI detectors estimate whether writing resembles model-generated text. The result should guide review, not replace context or human judgment.
Resources
Understand AI detection signals, confidence, passage-level evidence, limitations, false positives, and responsible review decisions.
Open core guideAI detectors estimate whether writing resembles model-generated text. The result should guide review, not replace context or human judgment.
Short samples, templated writing, translation, editing, and non-native writing can affect confidence. Good workflows document uncertainty.
No. AI detection estimates writing risk from text signals. It should be reviewed with context, drafts, citations, and policy before any high-stakes decision.
False positives can happen with short text, formulaic writing, templates, non-native writing, translation, or heavy editing. Review workflows should account for those limits.