Removing Claude's watermark in Word, Google Docs and email
Where hidden characters survive, where they get stripped for you, and how to clean text without breaking your formatting.
Removing the watermark · 3 min read
Where you paste Claude's output changes what survives. Worth knowing before you assume a document is clean, or assume it is not.
What survives a paste, by destination
| Destination | Hidden characters | Typographic punctuation |
|---|---|---|
| Word / Google Docs | Survive | Survive, and autocorrect may add more |
| Plain text editor | Survive | Survive |
| Email (rich text) | Survive | Survive |
A <textarea> on a web form | Survive | Survive |
| Retyping by hand | Destroyed | Destroyed |
The short version: pasting preserves everything. Only retyping removes it, and retyping a long document is not a plan.
Word and Google Docs make it worse, not better
Both have autocorrect rules that create typographic punctuation. Type two hyphens and Word gives you an em dash. Type a straight quote and it becomes curly. So a document can pick up exactly the markers people associate with model output purely from being edited in Word.
This cuts both ways:
- Your Claude text will not get cleaner by living in Word.
- Text you wrote entirely yourself in Word will contain the same em dashes and curly quotes, which is one reason typographic tells are weak evidence of anything on their own.
If you want ASCII output from Word, turn off autocorrect's smart-quotes and dash substitution before you start, not after.
The order that avoids losing formatting
The mistake is cleaning inside the formatted document, which either misses characters or wrecks the styling.
- Clean the text first, while it is plain. Paste Claude's output into a scanner, strip the characters, do the rewrite.
- Then paste the clean text into your document and apply formatting there.
Going the other way — formatting first, cleaning after — means either hand-editing inside Word (which will miss invisible characters, because you cannot see them) or copying out to clean and losing every heading, list and citation on the way back in.
Google Docs specifics
Docs does not show invisible characters and has no way to search for them by codepoint. Find-and-replace will not help: you cannot type a zero-width space into the search box in any reliable way.
So there is no in-Docs solution. Clean before you paste in.
If you already have a formatted Doc you suspect is dirty: File → Download → Plain text, scan that to find out what is in it, then decide whether it is worth rebuilding.
PDFs
A PDF is the worst case. Exporters routinely insert soft hyphens and zero-width characters around line breaks and hyphenation, so a PDF's extracted text is often full of them regardless of origin. Finding them in a PDF tells you little about whether the text came from a model.
If you are auditing, get the source document rather than the PDF.
What to actually do
Paste the text into the scanner while it is still plain, before it enters any editor. You will get exact counts of what is there. Strip and rewrite at that stage, then move the clean text into whatever document it needs to live in — and turn off your editor's smart-punctuation autocorrect so it does not undo the punctuation work.
Related: how to remove Claude's watermark · auditing a document · why AI text is full of em dashes