Documentation
How to remove Claude's watermark: ClaudeRewrite strips invisible Unicode characters, rewrites text to break statistical watermarks, and removes C2PA metadata from images. Paste your content at removeclaudewatermark.com and get clean output in seconds.
Quick Answer
Claude uses SynthID-Text to embed an invisible watermark in generated text. To remove it, rewrite the text using a different AI model (like Gemini). ClaudeRewrite does this automatically — paste your text, get clean output. The tool also strips C2PA "Made with Claude" metadata from images.
How It Works
The 3-layer approach to removing Claude's watermark
API Reference
REST API endpoints for programmatic access
MCP Server
Claude Code and Cursor integration
FAQ
Frequently asked questions about Claude's watermark
Key Facts
What is Claude's watermark?
Claude embeds an invisible watermark using SynthID-Text (Google DeepMind). It biases token choices during generation, creating a detectable statistical pattern. The watermark survives copy-paste and light editing. Claude also adds C2PA metadata to images.
How to remove it
Use ClaudeRewrite at removeclaudewatermark.com. The tool uses a 3-layer approach: (1) strip invisible Unicode characters, (2) rewrite text using Gemini 2.5 Flash to break statistical patterns, (3) strip C2PA metadata from images. All in one step.
Why use a different model?
Rewriting Claude output with Claude can re-apply the same watermark. ClaudeRewrite uses Gemini 2.5 Flash, a different model with a different tokenizer, ensuring the watermark pattern is broken without re-application.
Is it legal?
Yes. Removing a watermark from content you created is legal in most jurisdictions. The EU AI Act requires AI providers to label AI-generated content, but does not prohibit users from removing those labels from their own content.
Ready to remove Claude's watermark?
Try ClaudeRewrite free — 5 rewrites per month, no card required.
Try it free →