removellmwatermark.com

11 Aug 2026 · Anthropic confirms watermarking

Find the characters AI models hide in your text.

Paste anything written by Claude, ChatGPT, Gemini or Grok. Every invisible code point, look-alike letter and exotic space is listed with its Unicode name, and you get a clean copy back.

It runs in this tab. Nothing is uploaded, nothing is stored, and there is no account to make.

ScannerLocal · nothing leaves this tab
Cleaned

Every model, and what it puts in your text

Pick the model that wrote it. Each page covers what that specific version embeds, and what a cleaner can and cannot do about it.

Remove Claude Opus 5 watermarkZero-width characters · Variation selectors · Non-breaking and exotic spacesConfirmedRemove Claude Sonnet 5 watermarkZero-width characters · Variation selectors · Non-breaking and exotic spacesConfirmedRemove Claude Fable 5 watermarkZero-width characters · Variation selectors · Statistical (token-level) watermarksConfirmedRemove Claude Haiku 4.5 watermarkZero-width characters · Non-breaking and exotic spaces · Typographic punctuationLikelyRemove Claude Opus 4.8 watermarkZero-width characters · Typographic punctuation · Whitespace patterningLikelyRemove Claude Sonnet 4.6 watermarkZero-width characters · Non-breaking and exotic spaces · Typographic punctuationLikelyRemove Claude Code output watermarkZero-width characters · Variation selectors · Statistical (token-level) watermarksConfirmedRemove GPT-5.6 Sol watermarkZero-width characters · Typographic punctuation · Non-breaking and exotic spacesUnconfirmedRemove GPT-5.6 Terra watermarkTypographic punctuation · Non-breaking and exotic spaces · Whitespace patterningUnconfirmedRemove GPT-5.6 Luna watermarkTypographic punctuation · Whitespace patterningUnconfirmedRemove GPT-5.5 watermarkTypographic punctuation · Non-breaking and exotic spacesUnconfirmedRemove GPT-5.5 Pro watermarkTypographic punctuation · Non-breaking and exotic spacesUnconfirmedRemove GPT-5.4 watermarkTypographic punctuation · Whitespace patterningUnconfirmedRemove GPT-5.4 Pro watermarkTypographic punctuation · Whitespace patterningUnconfirmedRemove GPT-5.4 mini watermarkTypographic punctuationUnconfirmedRemove GPT-5.4 nano watermarkTypographic punctuationUnconfirmedRemove GPT-5.3 Instant watermarkTypographic punctuation · Non-breaking and exotic spacesUnconfirmedRemove GPT Image watermarkFile and document metadataConfirmedRemove Gemini 3.1 Pro watermarkStatistical (token-level) watermarks · Typographic punctuation · Non-breaking and exotic spacesLikelyRemove Gemini 3.6 Flash watermarkStatistical (token-level) watermarks · Typographic punctuationLikelyRemove Gemini 3.5 Flash watermarkStatistical (token-level) watermarks · Typographic punctuationLikelyRemove Gemini 3.5 Flash-Lite watermarkStatistical (token-level) watermarks · Typographic punctuationLikelyRemove Gemini 3 Pro watermarkStatistical (token-level) watermarks · Typographic punctuation · Non-breaking and exotic spacesLikelyRemove Gemini 2.5 Flash watermarkStatistical (token-level) watermarks · Typographic punctuationLikelyRemove Imagen watermarkFile and document metadataConfirmedRemove Grok 4.5 watermarkTypographic punctuation · Whitespace patterning · Non-breaking and exotic spacesUnconfirmedRemove Grok 4.3 watermarkTypographic punctuation · Whitespace patterningUnconfirmedRemove Grok 4.20 watermarkTypographic punctuation · Whitespace patterningUnconfirmedRemove Grok 4.1 Fast watermarkTypographic punctuationUnconfirmedRemove Grok 4 watermarkTypographic punctuation · Whitespace patterning · Non-breaking and exotic spacesUnconfirmedRemove Grok 3 watermarkTypographic punctuation · Whitespace patterningUnconfirmed

31 models · every page also available as markdown by adding .md to its URL

Questions

Asked most

What is an AI text watermark?
Two different things wear the same name. One is data hidden in the characters — invisible code points, look-alike letters, unusual spaces — which is exact and removable. The other is statistical: a key biases the model's word choices so the mark is spread through the writing itself, with nothing inserted. This tool removes the first and tells you honestly when it is looking at the second.
Did Anthropic really start watermarking Claude?
Yes. On 11 August 2026 Anthropic said every model released after 2 August automatically embeds an imperceptible watermark in generated text and files, to meet Article 50 of the EU AI Act — and applies it worldwide rather than only in the EU. It covers the Claude Platform API, claude.ai, Claude Code, and Claude models served through AWS, Google Cloud and Microsoft Foundry.
Does my text get uploaded?
No. The scanner is JavaScript running in your own tab. There is no request to clean, so there is nothing to log — which is why the free tier has no sign-in wall.
Is this legal?
Removing invisible characters from text you own is ordinary text processing. Whether you should present AI-written work as your own is a separate question, and it depends on the context you are submitting it to, not on the characters.