Article
Does Claude Opus 5.5 Watermark Its Text?
Yes. Anthropic's launch page says Claude Opus 5.5 comes with its watermarking measures. Here is what that means, what Anthropic has confirmed, and what it has not.
Yes, Claude Opus 5.5 watermarks its text. Anthropic's Opus 5.5 launch page says: "As with Fable 5.1, Opus 5.5 comes with our watermarking measures to comply with the EU AI Act." Opus 5.5 launched on September 22, 2026, after the August 2, 2026 cutoff that triggers watermarking.
The details below come from Anthropic's own pages. Where Anthropic has not said something, this post says so instead of guessing. For the quick per-model answer, see the Claude Opus 5.5 watermark page.
What Anthropic has confirmed
- Launch date: September 22, 2026.
- Watermarked: yes, per the launch page sentence quoted above.
- What the watermark is: Anthropic describes it in How Claude's text watermarking works. Future Claude models generate text that contains a watermark, and Anthropic says it is invisible to anyone without its detection API, has no practical impact on the quality or content of Claude's outputs, and contains no information about the user, their organization, or their conversations.
- Why: the EU AI Act. Anthropic signed the EU Code of Practice on AI-generated content in July 2026, and models launched on or after August 2, 2026 are marked from launch. Anthropic has since added the watermark to earlier models too, including Opus 5.
Can anyone detect Opus 5.5 text?
Not publicly. Anthropic's detection API is in private preview and limited to eligible organizations such as regulators, law enforcement, media, fact-checkers, independent researchers, educational organizations, and EU civil society groups. It is also available to enterprises that need to check their own Claude output. There is no public checker for text, and no third-party tool can read the watermark. Anthropic says it plans to expand access to the detection API over time.
What we cannot confirm
- How reliably the watermark can be detected in short passages. Anthropic has not published per-model figures that we could find.
- How much editing it takes for the signal to disappear. We do not claim a number.
- Whether a given rewrite removed the signal. Without a public detector, nobody outside Anthropic's program can measure it.
Updated November 12, 2026: an earlier version of this post listed Sonnet 5.5 and Haiku 5.5 as not confirmed. Anthropic's support page lists both as text-watermarked, along with the newer Fable 5.2 (see the watermark tracker).
What to do if you use Opus 5.5 for published writing
The watermark sits in the model's word choices rather than in a hidden character, so deleting characters does not remove it. A complete rewrite through a different, non-watermarking model changes the word-level pattern. That is what ClaudeRewrite does: it rewrites the wording with DeepSeek V4 Flash, keeps your facts and numbers, and also strips stray invisible characters and file metadata. Because there is no public detector, nobody outside Anthropic's program can verify the result either way, so treat it as risk reduction rather than a guarantee.
References
Get started
Ready to remove Claude's watermark?
Try ClaudeRewrite free — 5 rewrites per month, no card required.
Try it free