01
> source text
PASTE
Paste your text.
Bring a Claude response into the editor.
CLAUDECLEAN / TEXT IN → TEXT OUT
Remove Claude's watermark from your text.
Paste Claude's output. Scan and strip hidden characters locally, rewrite with a non-Claude model, copy the diff-checked result. Or upload .txt, .docx and .pdf files.
claudeclean.local
TEXT STAYS PRIVATE
01 / SOURCE TEXT
Paste text from Claude
Markdown and code blocks supported0 / 2,000 free chars
HOW IT WORKS
How to remove Claude's watermark in three steps.
See the signal, then choose what happens next.
02
scan report
3 zero-width1 soft hyphencleanable signals2SCAN & CLEAN
See what changed.
Find hidden characters and clean them locally.
03
before → after
clearusefulanswerfocusedhelpfulresponse
Copied ✓REWRITE & COPY
Change the pattern.
Rewrite, compare the diff, and copy the result.
WHAT YOU GET
What the Claude text cleaner does.
Everything stays focused on the text.
SCAN
scanZWSHOK
Find hidden characters.
Count hidden characters before you clean.
INTENSITY
lightmediumstrong
Choose the rewrite.
Pick light, medium, or strong.
DIFF
clearfocusedanswerresponse
Review every change.
See removals and additions side by side.
PRIVACY
Your text stays yours.
Scan runs locally. Text leaves only when you click Rewrite.
BACKGROUND
What is Claude's watermark?
SIGNALS
Two different signals
Claude-related watermark discussions usually involve two different signals. One is visible to a character scan: hidden Unicode such as zero-width spaces or joiners, soft hyphens, and letters from another script that resemble Latin characters. These can sometimes arrive when text is copied. The other is statistical: a pattern in word choices and token sequences, sometimes described as SynthID-style text watermarking. It is distributed through the writing itself, not stored in one character. A local scan can report the first kind, but it cannot see or measure the second. The distinction matters when choosing a cleanup step.
CLEANUP
What ClaudeClean removes
ClaudeClean's local scan finds supported hidden characters, strips them, and shows the count before you copy. If you choose Rewrite, the cleaned text is sent to a non-Claude model for up to three passes, aiming to change most words and sentence structures. The rewrite then removes typical AI-style habits such as em dashes, “not X but Y,” hedges, and summary closers. Keep em dashes is optional when source dashes should remain. The before/after diff is always shown, so you can inspect what changed instead of accepting an opaque result.
LIMITS
What it can't promise
No detector result is promised. Rewriting is best effort, and changing a statistical pattern can also change phrasing in ways that need human judgment. Review the meaning, names, numbers, links, and formatting before publishing. Scan and clean run entirely in the browser. When you choose Rewrite, the text is sent to OpenAI for processing and is not stored by ClaudeClean after the request. That boundary is why the tool shows the original and result side by side: you can decide whether the output is accurate and useful rather than relying on a label.
PRICING
More room to rewrite.
ClaudeClean Pro pricing: one simple plan. Cancel anytime.
PRO
CLAUDECLEAN PRO
ClaudeClean Pro
$5/ month
Cancel anytime.
- 300 credits every month (≈300,000 characters of rewriting)
- 20,000 characters per request
- Clean .txt, .docx and .pdf files
- Light / medium / strong intensity
- Dashboard with history
Checking account…
Free: 10 credits a month with an account, or try 2,000 characters right here without one. Strong rewrites cost 2× credits.
FAQ
Claude watermark FAQ.
No. Rewrite is a best-effort change, so review the diff and decide what to use.
It counts supported hidden characters, mixed-script lookalikes, and reports Style signals through an informational score and flag list.
Claude's output carries a statistical token-pattern signal (SynthID-style watermarking) plus recognisable stylistic habits. ClaudeClean rewrites your text with a non-Claude model in up to three passes until most words and sentence structures change, then strips typical AI-style patterns (em dashes, 'not X but Y', hedges, closers). It's best-effort: review the diff; no detector result is promised.
1 credit ≈ 1,000 characters; Light and Medium cost 1×, Strong costs 2×. Scan and hidden-character cleaning are free.
Scan and clean run in your browser. Rewrite sends text to our server and OpenAI for processing, then we discard it.
$5/month includes 300 credits per billing cycle, 20,000 characters per request, and cancel-anytime billing.
Upload .txt, .docx, or .pdf files; images are not supported. TXT and DOCX stay intact, while PDFs are regenerated from text.
Cancel in the billing portal. Email us within 7 days for a current-month refund request if you have not used credits.
READY WHEN YOU ARE
Clean Claude's watermark now.
See what changed before you copy.