Free forensics tool by ArticleGrade

Find, decode & remove the hidden AI watermarks in any text.

Paste text and see the exact machine fingerprints it carries — 150+ invisible Unicode codepoints, ChatGPT copy artifacts, homoglyph swaps, hidden zero-width messages and model-specific style tells. Evidence you can see and verify, not a black-box percentage.
✓ 100% in your browser ✓ No signup ✓ No word limit ✓ Results never paywalled ✓ No humanizer sold here
150+Codepoints scanned
5Model fingerprints
3Analysis modes
100%Local analysis
Scan any text for hidden AI watermarks
Paste it, drop a file, or fetch a live URL — one analysis checks 150+ invisible codepoints, copy artifacts, homoglyphs, style fingerprints, hidden messages and word-level predictability.
What you see
What's hidden
The hidden layer appears here live as you type — invisible characters, homoglyphs, AI-typography and AI-vocabulary surface in real time.
⇪ drop files here
or check a live page
0 characters · 0 words
Analysis runs locally in your browser — only the optional URL box contacts our server, to fetch a public page. Pasting or uploading text stays 100% on your device.
Honest limits: statistical watermarks — Google's SynthID-Text, OpenAI's unreleased scheme, and the watermark Anthropic began applying to Claude output in August 2026 under the EU AI Act — are embedded in word-choice probabilities and can only be verified by the provider holding the key. No third-party tool can detect or remove them, including this one (and including every "watermark remover" that claims otherwise). What can be reliably shown is deterministic character evidence and stylometric signals — treat the score as evidence, not proof, and never as grounds for an accusation on its own.
Why it matters

Text is no longer just what you can see.

Invisible characters, copy artifacts and model watermarks now ride along inside ordinary-looking paragraphs — marking where they came from, tracing who leaked them, or quietly betraying that "your" writing was machine-made.
U+202F

The narrow no-break space documented as a systematic ChatGPT (GPT-o3/o4-mini) insertion — invisible, but a machine-readable fingerprint.

Rumi, Apr 2025
Aug 2026

Anthropic began watermarking all Claude text output for EU AI Act compliance — applied globally, undetectable by any third party.

EU AI Act, Art. 50
61%

False-flag rate of AI detectors on non-native-English essays — why several universities disabled them and why evidence beats a percentage.

Stanford-lineage study
~50%

Of newly published web articles are now AI-generated — publishing more no longer sets you apart; verifiable quality does.

Graphite, 2025
Who uses this

One tool, many jobs — because the character layer is universal.

Whatever you're checking, the deterministic evidence is the same: exact characters at exact positions, shown so you can act on it — or defend against it.

Editors & publishers

Triage a slush pile in bulk. Flag pasted-from-ChatGPT submissions by their copy artifacts before a manuscript wastes staff time — with confidentiality, because nothing is uploaded.

SEO & content teams

Scan writer deliverables for watermarks and clean them before publishing, so hidden characters don't leak into your CMS or trip a client's detector.

Teachers & reviewers

Get per-sentence evidence you can actually show — a zero-width space at character 132, not an unfalsifiable percentage. With an explicit "inconclusive" verdict when the sample is too short.

Students proving their work

Self-check before submitting, see which lines trip detectors, and export an evidence report to pair with your document version history if you're wrongly accused.

Legal & compliance

Detect hidden zero-width payloads used to trace leaks, and audit documents for covert steganographic channels — entirely offline.

</>

Developers & security

Inspect user-submitted text for invisible-character injection, homoglyph spoofing and Unicode obfuscation before it hits your database or renders in your UI.

Writers & freelancers

Strip AI-tool artifacts (curly quotes, em-dashes, NBSPs, markdown) from work drafted with assistance, so your final copy is clean and portable.

Anyone pasting from AI

ChatGPT, Claude and Gemini output carries invisible baggage. One click shows it; one click removes it.

How it works

Three layers of evidence, in about a second.

Character forensics

Every character is checked against 150+ known invisible codepoints, exotic spaces, Cyrillic/Greek homoglyphs and ChatGPT copy artifacts — and any zero-width payload is decoded. These are exact, reproducible facts.

Stylometric signals

Sentence-length burstiness, lexical diversity, AI-vocabulary density and repetitive openers are measured, then matched against five models' known style fingerprints. Clearly labeled as probabilistic.

Evidence & cleanup

You get a tiered, sentence-level report you can export — and a configurable remover that strips the character-level watermarks while telling you honestly what cleaning can and can't do.

What we scan for

The full detection surface.

Most "watermark remover" tools strip a handful of characters and call it done. Here's everything this engine actually inspects.

Invisible characters

36

Zero-width spaces & joiners, word joiners, BOM, soft hyphens, bidi controls, Hangul fillers, tag characters.

U+200B · U+200C · U+2060 · U+FEFF · U+202E

Exotic spaces

15

Non-breaking, narrow no-break (the ChatGPT tell), en/em, thin, hair, figure and ideographic spaces.

U+00A0 · U+202F · U+2009 · U+3000

Homoglyphs

50+

Cyrillic and Greek lookalike letters substituted into Latin text — a deliberate evasion / watermark trick.

а е о р с · Α Β Ε Ο Ρ

Typography tells

6

Em/en dashes, curly quotes and ellipsis characters — weak on their own, telling in combination.

— – " " …

Chat artifacts

11

Phrases that only survive an un-edited copy-paste from an AI assistant.

"As an AI model" · "Sure! Here's" · "You're absolutely right"

Copy / markup artifacts

2

ChatGPT's data-start/data-end HTML attributes and raw markdown left from pasting.

data-start · **bold** · [x](url)

Stylometry

5

Burstiness, lexical diversity, AI-vocabulary rate, repetitive openers and em-dash density.

80+ AI phrase markers

Model fingerprints

5

Style signatures for the major assistants, scored side by side.

ChatGPT · Claude · Gemini · Grok · DeepSeek
How we compare

The evidence layer everyone else skips.

Feature and pricing data researched August 2026 from vendor pages and independent reviews.
ToolFree tierSignupCharacter forensicsHidden-msg decodeModel attributionEvidence shownPrivate (client-side)Sells humanizer
This toolUnlimitedNo✓ 150+ codepoints + HTML artifacts✓ zero-width steganography✓ all 5, evidence-listed✓ exact chars & positions✓ 100%No
GPTZero10k words/moFor dashboard~ flags "obfuscation"~ sentence highlights✗ retains textNo
Originality.aiNone meaningfulYes + payment~ separate tool~ per-sentence✗ stores scansNo
CopyleaksLimitedYes~ "AI Logic"✗ trains on your textNo
ZeroGPT~15k chars, adsNo✗ bare %~ paraphraser
TurnitinInstitution-onlyInstitution~ manipulation flags~ segments✗ retains papersNo
Pangram2k words/dayFor more~ sentence-levelNo
Unicode cleanersFreeNo~ strip & count✗ rarely✗ no explanation✓ mostly~ upsell funnels
One thing the classifier SaaS tools do that we deliberately don't: probabilistic "AI likelihood" verdicts from trained models (Pangram is the accuracy leader there). Our lane is verifiable evidence — the layer they almost all ignore. For a full quality read on whether Google and AI will surface your page, run it through ArticleGrade.

Clean text is only half the battle.

Removing watermarks makes a page tidy — but will Google rank it and will AI quote it? ArticleGrade grades any URL across six quality dimensions in ~20 seconds, free.

Grade my page free →
FAQ

Questions, answered honestly.

What is an AI text watermark?

A hidden signal marking text as machine-generated. Two kinds exist: character-level watermarks — invisible Unicode characters (zero-width spaces, special spaces, lookalike letters) inserted between normal letters — and statistical watermarks (Google SynthID-Text, Anthropic's Claude watermark since Aug 2026), which bias the model's word choices in a secret, key-verifiable pattern. This tool detects the first kind completely and flags the stylistic footprint that accompanies AI text generally.

How does the hidden-message decoder work?

Zero-width steganography hides binary data in invisible characters — for example ZWSP for a 0 bit and ZWNJ for a 1 bit. It's used for invisible document watermarking, tracing who leaked a file, and covert messaging. The decoder extracts every zero-width character in order, tries the common binary encodings, groups the bits into bytes, and shows any readable text it recovers. If a payload is present but not in a recognized text encoding, it tells you that too — all locally.

Does ChatGPT really watermark its text?

In April 2025, researchers documented GPT-o3/o4-mini systematically inserting the narrow no-break space (U+202F) into longer outputs. OpenAI called it a training quirk rather than a watermark and removed it within weeks — but text generated in that window carries it, and AI-pipeline tools still leave NBSP, zero-width and other artifacts. OpenAI has also built a true statistical watermarker (reported ~99.9% accurate) but has never released it or its detector.

Can any tool detect SynthID or Claude's new watermark?

No third-party tool can — including this one, and including every site claiming to "remove Claude's watermark" by stripping Unicode (that removes nothing: Anthropic's watermark, applied to Claude output since 2 August 2026 under the EU AI Act, lives in word-choice statistics, not characters). Detection requires the provider's key. Google's SynthID Detector only verifies Google-generated content; Anthropic says a public detector will come later.

Does removing invisible characters make AI text undetectable?

No — and any tool implying it does is misleading you. Stripping characters removes character-level watermarks and copy artifacts, which matters for clean documents and privacy. But statistical detectors (GPTZero, Originality, Turnitin) analyze word-choice patterns, which survive character cleaning entirely. We say this plainly because most "watermark remover" sites won't.

Can this prove someone used AI? Can it clear me?

Neither, on its own. Deterministic findings (invisible characters, chat artifacts) are strong evidence of machine processing but not proof of authorship; a clean result doesn't prove human authorship either. If you're accused: this tool's evidence report plus document version history (Google Docs, Word) is a far stronger defense than any detector percentage. Detection scores alone should never decide an accusation — several universities have disabled AI detectors over false positives.

Is my text uploaded anywhere?

When you paste or upload text, no — the analysis, decoding and cleaning all happen in your browser's memory with zero network requests (verifiable in your dev tools); reload and everything is gone. The one exception is the optional "check by URL" box: to fetch a page the browser can't reach directly, that public URL is sent to ArticleGrade's server, which retrieves the page and returns its text — then the watermark analysis still runs locally on your device. Either way, we never store your text or train on it, unlike mainstream detectors (at least one trains on your submissions by default).