Skip to main content

Guides

What Is an AI Text Watermark? Character-Level vs Statistical

CleanPastedText editorial · Updated

Text workbench

Try it on your text

Cleaning mode

Removes hidden characters and standardizes AI-style punctuation.

Try a real example

Each sample contains a problem you cannot see.

Original

Pasted text

0 chars · 0 words

Cleaned

Ready to copy

0 changes

Text stays in this browser

Same words · No AI rewriting · No content logging

What is a text watermark?

A text watermark is a deliberate, machine-detectable signal embedded in generated text to mark where it came from. Unlike an image watermark, it has to survive in plain characters — so every scheme in existence falls into one of two families, and the difference between them decides everything about detection and removal.

1. Character-level marks

Special Unicode characters hidden in the text itself: narrow no-break spaces (U+202F), zero-width characters (U+200B), tag characters (U+E0000–U+E007F), variation selectors. They travel with copy-paste, are trivially detectable in the bytes — and are deterministically removable.

2. Statistical watermarks

No characters are added. The model's word choices are subtly biased with a secret key, so the text itself carries a statistical signature only the key holder can test for. Character cleaning cannot touch this layer; only heavy rewording or translation degrades it.

Character-level watermarks and artifacts

The canonical episode is dated precisely. On April 16, 2025, OpenAI launched o3 and o4-mini. Within days, the academia startup Rumi reported that these models were embedding narrow no-break spaces (U+202F) where ordinary spaces belong in longer outputs like essays — a pattern absent from their GPT-4o tests, and removable with find-and-replace. OpenAI told Rumi the characters were not a watermark but "a quirk of large-scale reinforcement learning," and by late April they were gone from output. Whatever the intent, the episode proved the point: hidden characters do appear in AI text, they can fingerprint it, and knowing how to remove ChatGPT watermark characters is basic hygiene.

The same character layer supports true steganography. Unicode tag characters (U+E0000–U+E007F) encode an invisible copy of ASCII, and variation selectors (U+FE00–U+FE0F, U+E0100–U+E01EF) offer 256 invisible values — one hidden byte after any visible character, as Paul Butler demonstrated in his February 2025 post "Smuggling arbitrary data through an emoji". Anyone — a model vendor, a paste site, a colleague — could tag text this way, which is why a transparent character-level report matters. The complete invisible-characters list covers every code point in this layer, including the security research on each.

Statistical watermarks: SynthID-Text and Claude

A statistical watermark adds nothing to the text — it changes how the text is written. Google DeepMind's SynthID-Text, published in Nature in October 2024 (Nature 634:818–823), works by "tournament sampling": as the model picks each next word, candidate words compete in small tournaments scored by a function seeded with a secret key, and the winner is chosen. Repeated over a full answer, the output subtly prefers tournament-winning words — a signature invisible to readers but testable by anyone holding the key. Google live-tested it on roughly 20 million Gemini responses, has deployed it in Gemini since 2024, and open-sourced the scheme.

As of August 2026, Claude is watermarked the same way. On August 11, 2026, Anthropic announced that Claude models launched on or after August 2, 2026 embed a SynthID-Text-style statistical watermark in generated text, worldwide — plus C2PA provenance metadata for generated files, with a third-party detection API planned (TechCrunch, August 11, 2026). Anthropic is explicit that no Unicode characters are embedded, that detection needs longer samples, and that heavy editing or paraphrasing strips the mark.

OpenAI, meanwhile, built one and shelved it. The Wall Street Journal reported on August 4, 2024 that OpenAI had held back a ChatGPT text watermarker it judged about 99.9% effective, after roughly two years of internal debate — nearly 30% of surveyed users said they would use ChatGPT less. In its own explainer, OpenAI calls circumvention "trivial," naming translation, rewording, and even "asking the model to insert a special character in between every word and then deleting that character."

Why is everyone watermarking now? The EU AI Act

Because since August 2, 2026, EU law requires it. Article 50(2) of the EU AI Act obliges providers of generative AI systems to ensure outputs are "marked in a machine-readable format and detectable as artificially generated." The obligation took effect on August 2, 2026; under the May 2026 provisional agreement, systems already on the market get a grace period until December 2, 2026. Non-compliance carries fines of up to €15 million or 3% of worldwide turnover. Anthropic's August 2026 watermark launch, nine days after the deadline, is the clearest sign of where the industry is heading: statistical marking as the default, with character-level tricks explicitly avoided.

Can you remove a text watermark?

Honest answer: it depends entirely on which layer the mark lives in.

Character-level: yes

Hidden Unicode is deterministic data, so removal is deterministic too. The AI text cleaner strips special spaces, zero-width characters, tag characters, and directional controls in your browser and names every character it found by code point — your visible wording stays byte-for-byte identical.

Statistical: no character tool can

SynthID-style marks are the word choices. No character cleaner — ours included — can remove them, and any tool claiming otherwise is misleading you. They degrade only under substantial rewording or translation, and short or fact-constrained text barely holds a mark in the first place.

Transparency-first cleaning still matters in the statistical era: it strips the trackable character layer (privacy), makes AI text behave like typed text in editors and CMSs (hygiene), and — because it shows you exactly which characters were found — tells you the truth about what is and is not in your text. For the layers of baggage that actually ride along in AI output, beyond the watermark question, see what's inside AI text.

Common questions

Frequently asked questions

Does ChatGPT watermark its text?

Not with a known watermark as of August 2026. In April 2025, ChatGPT's o3 and o4-mini models briefly embedded narrow no-break spaces (U+202F) in longer outputs; OpenAI said it was a quirk of large-scale reinforcement learning, not a watermark, and it disappeared within days. Separately, the Wall Street Journal reported in August 2024 that OpenAI built a ~99.9%-effective statistical watermarker for ChatGPT and shelved it after user-research pushback.

Does Claude watermark its text?

Yes. Anthropic announced on August 11, 2026 that Claude models launched on or after August 2, 2026 embed a SynthID-Text-style statistical watermark in generated text worldwide, plus C2PA metadata for generated files. Anthropic says no Unicode characters are embedded, detection needs longer samples, heavy editing or paraphrasing strips it, and a third-party detection API is planned.

Can a tool remove an AI text watermark?

It depends on the kind. Character-level marks (hidden Unicode such as U+202F or zero-width spaces) are deterministic and removable — CleanPastedText strips them in your browser and reports each one by name and code point. Statistical watermarks like SynthID-Text live in the word choices themselves, so no character-level tool can remove them; only substantial rewording or translation degrades them.

What is SynthID-Text?

SynthID-Text is Google DeepMind's statistical watermark for AI text, published in Nature in October 2024. During generation it runs secret-key-seeded 'tournaments' between candidate next words, subtly biasing which words are chosen. The pattern is invisible to readers, adds no characters, was live-tested on roughly 20 million Gemini responses, and has been deployed in Gemini since 2024.

Is AI text watermarking required by law?

In the EU, yes. Article 50(2) of the EU AI Act requires generative AI outputs to be marked in a machine-readable format and detectable as artificially generated, effective August 2, 2026 — systems already on the market have until December 2, 2026 under the May 2026 provisional agreement. Fines reach €15 million or 3% of worldwide turnover.

Do AI watermarks survive copy-paste?

Character-level marks travel with the text through copy-paste — that is what makes them useful as fingerprints and easy to strip with a cleaner. Statistical watermarks survive copy-paste, character cleaning, reformatting, and light edits, because the signal is the word choice itself; they weaken only under heavy paraphrasing or translation, and short or highly factual text barely holds a mark at all.

Continue reading

Related guides & tools

SynthID-Text Explained

SynthID-Text biases word choice with a secret key instead of adding characters. How tournament sampling works, who uses it, and why no cleaner can remove it.

Remove ChatGPT Watermarks & Hidden Characters

Remove ChatGPT watermark characters like U+202F and zero-width Unicode in one click. See exactly what was found — free, private, 100% in your browser.

AI Watermark Checker: Is Your Text Watermarked?

Check text for hidden AI watermark characters — U+202F, zero-width, tag characters — with a named per-codepoint report. Free, instant, nothing uploaded.

EU AI Act Watermarking Rules: What Article 50(2) Requires

What EU AI Act Article 50(2) requires for AI-generated content as of August 2, 2026 — and why removing hidden Unicode characters doesn't satisfy it.

C2PA Content Credentials Explained

What C2PA Content Credentials are, how the signed manifest works, and why a screenshot or social-media upload destroys them. With primary sources.

Invisible Unicode Characters: The Complete List

The complete table of invisible Unicode characters: zero-width spaces, U+202F, BiDi controls, tag characters, variation selectors — and why each one matters.

How CleanPastedText Works

The exact pipeline CleanPastedText runs: detection-first scanning, emoji-safe stripping, NFKC normalization. Deterministic, client-side, nothing uploaded.

How to Clean Text Copied From Claude

Claude output pastes with markdown symbols, curly quotes, and stray spacing left over. Clean it in your browser — free, and honest about the 2026 watermark.

How to Clean Text Copied From Gemini

Gemini output pastes with markdown symbols, dense tables, and stray spacing left over. Clean it in your browser — free, and honest about SynthID-Text.

AI Text Cleaner: Fix ChatGPT, Claude & Gemini Output in One Click

Free AI text cleaner: strip hidden Unicode (U+202F), smart quotes, and em dashes from ChatGPT, Claude, or Gemini text. 100% in-browser, nothing uploaded.