Does Claude Watermark Text? How It Works
Summary: Claude's Watermark in 30 Seconds Claude models released on or after August 2, 2026 embed an imperceptible mark in the text they generate. Older models are being retrofitted. The mark lives in the text itself. It survives copy and paste, and can survive light editing. Files get a separate mechanism: .svg, .png and .jpg outputs carry signed provenance metadata following the C2PA standard. Coverage is broad: the API, Claude apps, Claude Code, Cowork, Claude Tag, plus access through AWS, Google Cloud and Microsoft Foundry, worldwide. A detected mark means the content may have been processed by Claude. The absence of a mark proves nothing about whether text was AI-generated. Detection tooling and technical documentation are not published yet. Today, no third party can actually verify a mark. Watermarking an image is easy. You nudge millions of pixel values by amounts no eye can catch, and you are done. Watermarking text is a different problem entirely. There are no pixels, only words, and you cannot change a word “slightly”. It is either that word or a different one. ...