Clean supported Gemini text marks and review the result before copying.
The tool removes supported hidden characters and offers local statistical processing. SynthID Text changes token probabilities during generation, so deleting invisible characters alone is not proof of removal and results remain best-effort.
This page is for Gemini text. Text signals, hidden characters, and token-level marks are different mechanisms and should not be presented as one verified detector.
As Google changes SynthID coverage or exposes additional verification methods, the capability will be added here rather than moved to a new status page.
Separate model-level text watermarks from hidden characters, metadata, media labels, and interface formatting.
Use deterministic rules and show the result instead of asking another language model to regenerate the text.
Only call a method supported after repeatable tests identify the affected models, products, languages, and text lengths.
Google confirms SynthID Text for text generated in the Gemini app and web experience.
No. The public tool can process supported text signals, but it cannot verify Google's private production detector.
Not by itself. SynthID modifies token selection during generation, so any statistical processing is best-effort.