#1 Claude watermark removal - Text, images & files

Remove Claude's watermark. Keep what's yours.

Every new Claude model weaves a machine-readable pattern into the text it generates, and it survives copy-paste and light editing. Our structural rewrite breaks that pattern at the sentence level, while your meaning, facts, and code pass through untouched.

Free to tryNothing stored3-layer protectionC2PA metadata stripped

Input0 words

or drop a file here

Output

Text is sent only to the selected model. Facts, code and URLs are preserved. Only wording and rhythm change.

Why clean it now? No shipping Claude model is watermarked yet — the mark applies to models released from August 2, 2026 onward, and Anthropic is still rolling it out to older ones. But there is no public detector today, and the day Anthropic's detection API ships, anything you have already published becomes checkable retroactively. Cleaning your text now closes that window before it opens. See which models are marked →

Watermark rewrite

Break the statistical watermark — keep every fact.

ClaudeRewrite restructures Claude's output at the sentence level using a non-watermarking model, so the SynthID pattern is gone while your meaning, numbers, and citations stay exactly as you wrote them.

Try it free

Music has long been a universal language, shaping emotions, cultures, and personal identities across generations. Its rhythms and melodies transcend barriers, uniting people in ways that words alone rarely achieve.

Facts kept

meaning preserved

Every number & date preserved

Names and citations untouched

Statistical watermark reset

Meaning check

See exactly what changed before you publish.

Every rewrite ships with a meaning report that flags any number, date, name, or claim that shifted — so you can trust the clean version without re-reading it line by line.

Create a free account

0 changed

facts drifted

"2026 revenue rose 14%" — kept

"Dr. Alvarez (2021)" — kept

1 phrase reworded — flagged to review

Image metadata

Strip “Made with Claude” from images too.

Claude marks generated images with C2PA provenance metadata. ClaudeRewrite removes it in one pass, so exported images carry no hidden “generated by AI” signature.

Clean an image

C2PA gone

metadata stripped

c2pa.manifest removed

EXIF “software” tag cleared

Pixels untouched

API & MCP

Remove watermarks right inside your stack.

Ship clean text programmatically. Call the REST API from any language, or drop the MCP server into Claude Desktop, Cursor, or Claude Code so your agent can strip the watermark mid-workflow. Bring your own key for unlimited use at raw model cost.

One POST request → clean rewrite back as JSON
MCP server works with any MCP-compatible client
Same word quota as the app, or BYO OpenRouter key

Read the API docs Set up the MCP

terminal

# REST API — clean any Claude draft
curl -X POST https://removeclaudewatermark.com/api/v1/rewrite \
  -H "Authorization: Bearer YOUR_KEY" \
  -d '{"text":"Your Claude draft…"}'

# MCP server — for Claude Desktop / Cursor / Code
npx claudeclear-mcp
→ tool: remove_claude_watermark(text)

Your dashboard

One clean workspace for everything

Track usage, review every rewrite with its meaning-check, manage your API keys and team, and watch your affiliate earnings — all in one place.

removeclaudewatermark.com/dashboard

ClaudeRewrite

Overview

Remove watermark

Check risk

Strip image data

Billing & usage

API keys

Overview

Professional plan · resets Sep 1

New rewrite

Words this month

42,180

of 75,000

Rewrites

318

this month

Credits

6,000

non-expiring

Referrals

50% recurring

Recent activity

Document	Words	Check
Q3 board summary	1,240	Passed
Blog draft — rewriting vs synonyms	890	Passed
Client proposal v2	2,010	1 flag
Newsletter — August issue	1,530	Passed

Create your free dashboard

5 / month

free rewrites, no card, no signup

0 stored

text is processed and deleted, never logged

Instant

fast rewrite on a non-watermarking model (DeepSeek V4 Flash)

100% facts

meaning check verifies numbers, names and code survive

Three layers of protection. One tool.

Layer 1: Break the statistical watermark

This is the real watermark. Claude's mark lives in the model's word choices, so ClaudeRewrite rewrites your text at the sentence level to change those choices until the token pattern no longer matches. Facts, numbers and code stay intact.

Layer 2: Strip stray invisible characters

AI text copied out of chat interfaces often carries zero-width spaces, joiners and smart-quote artifacts. These are not the watermark, but they are common AI tells, so ClaudeRewrite removes them too. Deterministic and lossless.

Layer 3: Remove C2PA metadata

Strip 'Made with Claude' provenance from images, PDFs, and files. Your content shows no AI provenance.

Same meaning. Zero watermark.

Claude's output: statistically marked

It is important to note that, in the context of our ongoing efforts to enhance operational efficiency, we have implemented a comprehensive strategy designed to optimize resource allocation across all departments, thereby ensuring that we are able to maximize productivity while simultaneously minimizing unnecessary expenditures.

After ClaudeRewrite: clean

We cut waste by giving each department exactly the resources it needs. A single new strategy now steers that allocation, and it's already showing.

Simple pricing

Free to try. Pick a plan when you need volume.

Anonymous

$0 no signup

Try it now, no account.

3 rewrites / month
Up to 2,000 words each
All rewrite models

Rewrite text

Free account

$0 register

5 rewrites / month
Up to 2,000 words each
+1 rewrite per friend referred

Create free account

save 50% annually

Starter

For steady monthly use.

$9.99/month

25,000 words / month · unused words roll over

25,000 words / month, unused words roll over
Up to 5,000 words per rewrite
Meaning check on every rewrite
Rewrite history (30 days)
Email support

Professional

For writers and small teams.

$19/month

75,000 words / month · unused words roll over

75,000 words / month, unused words roll over
Up to 10,000 words per rewrite
API access
Bring your own API key for unlimited use
Priority models
Priority support

Business

For agencies and publishers.

$49/month

300,000 words / month · unused words roll over

300,000 words / month, unused words roll over
Up to 10,000 words per rewrite
API access
Team members (linked users)
Priority support

Enterprise

For high-volume teams. Buy online, no sales call.

$199/month

2,000,000 words / month · unused words roll over

2,000,000 words / month, unused words roll over
Up to 25,000 words per rewrite
API access + bring your own key
Team members (linked users)
Priority models and support
Invoicing available on annual plans

1 rewrite = one paste of up to 2,000 words free (10,000 words on paid plans). Unused monthly words roll over, so you never lose what you paid for.

Just need this one?

$9 for 30,000 word credits. No subscription, never expire.

Affiliate program

Earn 50% recurring, forever

One of the most generous commissions in the space. Refer anyone to ClaudeRewrite and earn 50% of every payment they make — every month, with no cap. Free to join. Move the slider to see what it adds up to.

Join the program

Earnings calculator

Active referrals10

1100

Per month

$95

Per year

$1,140

10 referrals on Professional ($19/mo) × 50% = $9.50/mo each, recurring for as long as they stay subscribed.

Straight answers, no fine print.

What is Claude's watermark?+

An invisible, machine-readable pattern Anthropic embeds in generated text to comply with the EU AI Act's transparency rules. It's baked into the model's word choices as the text is generated, not appended afterwards, so it travels with the text through copy-paste and can survive light editing.

Does my Claude, Claude Code, or Claude API output carry it?+

If it came from a supported Anthropic model, assume it does. The mark applies across the Claude platform, the API, Claude Code, and Claude Cowork, worldwide, including models served through AWS, Google Cloud, and Microsoft Foundry.

Does editing or rewriting remove it?+

Not reliably. Anthropic notes the signal can be lost under heavy edits, sentence-level paraphrasing, translation, and short passages, but a missing watermark isn't a guarantee the text is undetectable. ClaudeRewrite targets exactly that mechanism: a structural rewrite that changes the pattern itself, rather than hoping a few swapped words are enough.

How does ClaudeRewrite remove the watermark?+

Since the mark lives in the model's word choices, the only reliable fix is changing enough of them to break the pattern. Your text is sent to the model you choose on OpenRouter with a strict edit instruction: preserve every fact, number, name, code block and URL, but rebuild wording and sentence rhythm. The output no longer statistically matches the input's token pattern.

Is this guaranteed to pass every AI detector?+

No. And don't trust anyone who promises that. Detection isn't a binary, and a clean rewrite is not a certificate of origin. What a structural rewrite does is remove the specific statistical fingerprint Claude's watermark relies on. For legitimate work you own, that's the whole point.

Do you store my text?+

No. Your text is sent only to the model you pick, processed, and deleted. We never store, log, or use your writing for training.

What does it cost?+

Free to try: 3 rewrites a month, no card, up to 2,000 words each. Register for 5 a month, and earn an extra rewrite for every friend who signs up with your link. Paid plans start at $9.99 for 25,000 words a month, and unused words roll over to the next month instead of expiring.

Remove Claude's watermark from everything

Text, images, PDFs, and files. All three layers of protection in one tool.