What the tool can prove
AI Watermark Detector finds concrete Unicode signals: invisible characters, bidirectional controls, special spaces, and mixed-script identifiers. These are facts that can be verified by anyone with the same input.
What the tool cannot prove
The tool cannot determine whether a text was written by a human or an AI. It cannot detect statistical watermarks like SynthID Text, image pixel watermarks, or C2PA metadata. It also cannot access any model provider's private watermark keys.
Do not rely on it for legal decisions
Do not use AI Watermark Detector results as the sole evidence in legal, academic, or employment disputes. A finding of hidden characters is a signal to review, not proof of wrongdoing.
Always review before cleaning
The cleaning feature removes characters permanently from the output. Make sure you understand what is being removed before copying the cleaned text.