Read your watermark report
What this finds in Claude text
This tool finds observable text artifacts such as zero-width characters, directional controls, non-standard spaces, and control characters. Those can come from many editors or workflows, so they are clues rather than a Claude fingerprint.
Anthropic’s newly documented Claude mark is woven into model output and is not the same as a hidden Unicode character. Anthropic has not published the private detector needed to verify that statistical mark from pasted text.
Why AI detection needs caution
The style section flags a few formulaic phrases or unusually uniform sentence patterns. Human writing can contain them, and AI writing can avoid them. This is not a probability of authorship and should not be used as a high-stakes verdict.
- A Claude mark can mean Claude processed or edited text.
- A missing mark does not prove that AI was not involved.
- A hidden character can come from copy and paste, not only AI.
Remove visible text artifacts
If you are preparing text for a CMS, form, or document, copy the cleaned version after reviewing each finding. Cleaning observable characters cannot remove Anthropic’s statistical watermark, and it does not change the provenance of the ideas.
When you publish the result on a public page, check whether the crawlers you choose to allow can reach it.