Anthropic now watermarks Claude's text output — here's how to remove it

Claude models released on or after August 2, 2026 embed an invisible statistical watermark in generated text, and Anthropic is extending it to older models. These hidden patterns can flag your content as AI-written. This tool works on both layers: Quick Clean strips 34+ invisible Unicode characters — 100% in your browser, nothing uploaded; Deep Rewrite uses LLM rewriting to break statistical patterns (best-effort). Free, no account needed.

The tool supports a generous input limit for Quick Clean and a substantial limit for Deep Rewrite. For users needing more Deep Rewrites, a referral system is available where sharing a personal link grants extra credits.

Key features include:

• Free and unlimited Quick Clean

• Daily free credits for Deep Rewrite

• Runs entirely in the browser for Quick Clean

• No account or sign-up required

• Supports multiple AI models

This utility is designed to help users maintain greater control over their AI-assisted content, whether for academic, professional, or publishing purposes.