Simple Unmark guide
How to Remove AI Watermarks From Text
- Published
- Last updated
- Sources checked
- 3 primary references
Short answer
Work out which mechanism applies first. Hidden characters are removed exactly by a scanner. A statistical token-choice watermark needs the wording genuinely rewritten — providers state that light edits and synonym swaps do not work, while a complete rewrite does. Then verify facts, and never treat a detector score as proof either way.
Hidden characters and statistical watermarks need different treatment, so a complete pass addresses them in order rather than hoping one tool covers both.
1. Identify the mechanism
Start here, because it determines whether the job is exact or probabilistic. Text from a provider that documents a token-choice watermark — Gemini, and Claude since August 2026 — carries a statistical signal and no added characters. Text from a provider with no documented text watermark may still carry stray invisible characters from the copy path.
Our provider status tracker records what each provider documents, with sources and a verification date, so you are not guessing.
2. Preserve the original
Keep an untouched copy before you change anything. You will need it to compare meaning, numbers, names, formatting and citations afterwards — and a semantic rewrite is precisely the kind of edit that can quietly alter a figure or a hedge.
3. Run deterministic cleanup
Scan for zero-width spaces, direction controls, soft hyphens, tag characters and unusual Unicode spaces, and remove them. This part is exact: a good tool tells you which code points it found and how many.
Be conservative. Do not strip zero-width joiners and non-joiners by default, because they are essential to emoji sequences and to scripts including Arabic, Persian, Hindi and Malayalam. Do not "normalise" every non-ASCII character, which mangles legitimate text.
4. Rewrite the statistical layer properly
Change syntax, sentence boundaries, clause order, transitions and vocabulary across the whole passage. A synonym pass is not enough: Google describes the method as robust to modifying a few words and mild paraphrasing, and Anthropic contrasts light editing, which probably will not remove the mark, with a complete rewrite, which will.
Note the honest limits. Short passages give a rewrite little room. Dense factual text and code constrain how much can legitimately change — and they also carry less watermark to begin with, which cuts the same way.
5. Verify, and calibrate your claims
Compare every factual claim, number, name and quotation against the original. For high-stakes content, have a qualified human review it.
Then be realistic about what you have achieved. For most providers no public detector exists, so nobody — including us — can measure the residual signal. You have reduced a signal, not proven its absence, and a detector score is not proof of authorship in either direction.
Sources
- SynthID text watermarking and detection
Google AI for Developers · first-party
- How Claude's text watermarking works
Anthropic · first-party
- UAX #15: Unicode Normalization Forms
Unicode Consortium · first-party
Published by Simple Unmark. Our methodology page explains how claims on this site are sourced, what we test, and what we refuse to claim.
Apply this to your own text.
Three guest cleans are free, with no account required. Deterministic character inspection is free and unlimited.
Keep reading
- How AI Text Watermarks WorkHow hidden-character marks and statistical token watermarks differ, how detection works, and why removal claims should always be bounded.
- What Is SynthID Text?SynthID Text explained: how Google's text watermark is applied during generation, how detection works, what weakens it, and how it differs from image watermarking.
- Hidden Unicode vs Statistical Text WatermarksThe two mechanisms confused most often: invisible Unicode characters versus keyed token-choice watermarks. What each one is, what removes it, and what neither proves.
- AI Watermarks vs AI Detectors vs Provenance MetadataThree different things routinely confused: keyed watermarks added at generation, statistical AI detectors that guess, and C2PA metadata attached to files.
