List your AI tool on AI Orion for free. Submit a tool

Claude Watermark (claudewatermark.xyz)

Paste your Claude text. Strip hidden HTML, invisible Unicode and copy-paste fingerprints free, then humanize the word-choice pattern in one click.

Pricing
Freemium
Starter, Pro, Max

About Claude Watermark (claudewatermark.xyz)

Claude Watermark is a free browser tool that checks pasted text for the traces Claude's chat interface leaves behind, such as hidden HTML class names, invisible Unicode characters and stylistic tells like curly quotes and repeated em dashes. It is aimed at people who copy Claude's output into another editor, including writers posting fiction on AO3 and anyone pasting into Google Docs, WordPress, Notion or similar platforms, and who want to avoid being flagged as having used AI.

Checking runs entirely in the browser, so no text is uploaded, and it is free and unlimited without an account. The tool looks for HTML class names containing claude, zero-width and invisible Unicode characters, byte-order marks, non-breaking and exotic spaces, plus stylistic signals such as curly quotes and em dash frequency. Deterministic findings come back with a count and a position within the passage; a paid rewrite pass then resamples every sentence through a different model.

Claude Watermark distinguishes deterministic findings, which are either present or not, from stylistic findings such as em dashes and curly quotes, which prove nothing on their own; measuring ten pre-computer novels found em-dash rates swinging from zero per thousand words in Austen and Stoker to twenty-six in Melville. Claude Watermark states plainly that no tool, including its own, can verify Anthropic's statistical watermark without its unreleased key, and keeps its detection engine open source so that claim can be checked.

Checking and cleaning are free, unlimited and need no account, while the rewrite engine runs on credits bought through monthly plans: Starter at $9/month for up to 100,000 words, Pro at $19/month for up to 300,000 words, and Max at $49/month for up to 1,000,000 words, alongside one-time top-up packs. Monthly credits stack rather than reset, one-time credits never expire, failed jobs are never charged, and text processed for a rewrite is not stored.

Use cases

  • Avoiding an AO3 Claude-detector flag on fan fiction
  • Cleaning Claude's HTML and Unicode traces before pasting into platforms that keep markup
  • Changing wording so an AI detector scores it as less predictable

Key features

  • Detects HTML class names Claude's chat interface leaves in pasted text
  • Detects zero-width characters, byte-order marks and exotic spaces
  • Flags stylistic tells like curly quotes and repeated em dashes
  • Removes detected artifacts from text in one click
  • Rewrite pass resamples every sentence through a different model
  • Runs checking entirely in the browser with no upload
  • Open-source detection engine
  • Keeps a live list of which Claude models are watermarked

Pros and cons

Pros

Checking and cleaning are free and require no account
Purchased credits never expire

Cons

Rewriting and cleaning the full wording require paid credits
No public tool, including this one, can verify Anthropic's official watermark

Pricing

Starter
$9/month
  • Humanize up to 100,000 words
  • 200 credits added monthly
Pro
$19/month
  • Humanize up to 300,000 words
  • 600 credits added monthly
  • Most popular
Max
$49/month
  • Humanize up to 1,000,000 words
  • 2,000 credits added monthly

Reviews

Loading reviews…

Frequently asked questions

What exactly does this remove?
Every trace actually detectable in the text: Claude's HTML class names, zero-width and invisible Unicode characters, non-breaking and exotic spaces, and smart quotes and em dashes, plus a rewrite pass that varies sentence rhythm and length.
Is it free?
Checking is free, unlimited and needs no account, and runs in the browser; cleaning and rewriting cost credits, which never expire.
Will my text pass an AI detector?
The rewrite pass varies sentence length, rhythm and phrasing, which moves the predictability signal AI detectors measure; copy-paste artifacts are removed outright.
How does the AO3 Claude detector work?
It is a single CSS rule matching paragraphs whose class name contains 'claude' inside a work's body, present only if the text was pasted straight from Claude.
Is my text private?
Checking runs entirely in the browser and is never uploaded; a rewrite sends the passage to the model provider and it is not kept afterward or used for training.
Does Claude watermark its text?
Yes, on supported models; Anthropic lists Claude Fable 5.1 and Mythos 5.1 as watermarked at launch, with earlier models in a transition period.

Alternatives to Claude Watermark (claudewatermark.xyz)

More in AI Detector