Skip to content

Documented as of August 2026

Claude Watermark Remover for Text

Anthropic documents a key-based watermark woven into Claude's word choices, and states that nothing is added to the text and there are no hidden characters. Because the mark lives in the wording, a thorough rewrite is what affects it. Simple Unmark rewrites the passage and removes any stray characters from the copy path.

Privacy mode

Anthropic Claude: current status

· Documented by the provider

Anthropic documents a text watermark in Claude output, applied globally, using a SynthID-Text-style token-choice method.

Status
Rolling out
Mechanism
A key-based pattern in low-stakes token choices. Anthropic describes using the key plus the preceding words to settle which word the model picks, and states the approach is a version of the SynthID-Text method published in Nature.
Hidden characters involved?
No. Anthropic states plainly that nothing is added to the text and that there are no hidden characters.
Public detector
Anthropic offers its watermark detection API in restricted private preview to eligible organizations and enterprises with EU compliance obligations. It is not a general public checker, and Simple Unmark does not have access.

What the provider says weakens the signal

  • Anthropic states light editing probably will not remove it, while a complete rewrite in which every word is replaced will
  • Text Claude only lightly edited, because few of the words are Claude's
  • Factual passages and code, where there are fewer free choices to encode
  • Small samples, where detection does not work well

How this works

Deterministic marks can be removed exactly. Probabilistic patterns can only be reduced. Simple Unmark keeps the two separate and does not promise detector outcomes.

  • Anthropic describes the watermark as using the key plus the preceding words to settle which word the model picks, and identifies it as a version of the SynthID-Text approach published in Nature in 2024. It applies globally rather than only in the EU.
  • Anthropic says light editing probably will not remove the watermark, while a complete rewrite in which every word is replaced will. Swapping a few words leaves most of Claude's choices in place.
  • Anthropic also notes the mark is sparser where Claude had fewer free choices, such as factual passages and code, and that detection does not work well on small samples. Text Claude only proofread carries very little, because nearly all the words are the person's.

Three steps

  1. 1

    Paste the Claude text. The local scan reports any invisible characters, which come from the copy path rather than from the watermark.

  2. 2

    Run the clean. The rewrite is the step that matters for this watermark, because it replaces Claude's word choices throughout the passage.

  3. 3

    Check the rewrite kept your facts and figures, then copy it.

Worked example

Before

The library will extend its opening hours from 1 March. On weekdays it will stay open until 9 p.m., and on Saturdays until 6 p.m. The change follows a survey in which 68% of members asked for later hours.

After

Starting 1 March, members get later opening times. The doors now close at 9 p.m. Monday to Friday and at 6 p.m. on Saturday, after 68% of surveyed members said they wanted the library open later.

Illustrative editing example written by us. The date, times and percentage stay fixed because changing them would change the facts; the words around them are new. Anthropic notes that fixed facts like these carry less watermark anyway. Not tested against a Claude detector, which is not public, and real output varies per request.

Scope

What this page can and cannot do

Can

  • Replace Claude's word choices across the whole passage, which is the change Anthropic describes as removing the mark
  • Clean invisible characters that entered through a browser, editor, CMS or collaboration tool
  • State the current status with a first-party citation and a verification date

Cannot

  • Check for the watermark. Anthropic offers restricted detection API access in private preview, but Simple Unmark does not have access
  • Guarantee that every trace is gone, since we cannot measure what we cannot detect
  • Strip C2PA provenance metadata from files. This tool handles pasted text, not file metadata
  • Tell you whether a given passage was written or only edited by Claude; Anthropic notes the mark cannot distinguish those either

We don’t store submitted content or cleaned results in our application database, and we don’t send content to analytics. For rewrites, Private mode routes plaintext through the app, its load balancer and the processors. Confidential mode encrypts text from the browser to the workload, so the web backend only receives metadata. DeepInfra still rewrites the text. The deterministic scan in Invisible character remover runs entirely in your browser. See Privacy Policy and Methodology.

Questions

About Claude Watermark Remover for Text

Sources

Does Claude watermark text?
Yes, per Anthropic's own documentation published in August 2026: supported Claude models weave a machine-readable watermark into generated text, applied worldwide. Models launched before 2 August 2026 were to be updated over the following months. See the dated status block above.
Is it a hidden Unicode character I can strip?
No. Anthropic states directly that nothing is added to the text and there are no hidden characters. Tools that claim to strip a Claude watermark by deleting invisible characters are not addressing this mechanism.
Does the watermark survive copy and paste?
Yes. It is part of the word choices themselves, so copying, pasting and reformatting carry it along. Anthropic says light editing probably will not remove it.
Does the watermark prove Claude wrote something?
No, and Anthropic says so explicitly: it can only indicate Claude was likely involved at some point, and cannot separate "Claude wrote this" from "Claude heavily edited this". It says nothing about ownership or authorship.