All platforms
Does Claude watermark the text it writes?
Soon, and you will not be able to see it:
- Anthropic says future Claude models will write text with a hidden watermark, to comply with the EU AI Act.
- Nothing is added to the text. There are no hidden characters, and readers cannot tell the difference.
- The watermark carries no identifying information. It cannot be traced to a person, a company, or a chat.
- It is detectable by anyone with the key. Anthropic is offering a detection API in private preview to regulators, media, researchers, and companies that must check watermarks under EU law.
- Detection is weak on short text and on light edits. If Claude only proofreads your writing, there is little for the watermark to attach to.
What this means for marketers
No action required, but the ground has moved: text Claude drafts heavily can now be flagged as likely Claude, while text it only polishes mostly cannot. Worth considering that difference before calling a piece “human-written.”
Primary source Anthropic: how Claude’s text watermark works (August 14, 2026)
Links checked September 25, 2026
How we check this. *Beep Boop Bop 🤖 Never ever, ever legal advice.