An AI signal is a starting point.
This product does not claim a universal detection accuracy or an ability to prove authorship.
Understand the unit
A provider’s 0–1 probability is a model signal. It is not a percentage of AI-written words, nor is it interchangeable with another provider’s score.
Leave uncertainty visible
Short, unsupported, ambiguous, or incomplete input may not support a useful result. Missing scores stay missing; the application never substitutes a local keyword classifier.
Use more than one source of evidence
Consider draft history, sources, author context, and the relevant policy. A detection score should never be the sole basis for a consequential decision.
Quality evaluation is ongoing
Provider availability does not establish product-level accuracy. Evaluation needs representative human, AI-assisted, and generated text, including non-native English writing.