Remove Watermark
Find text watermarks repeated across your PDF's pages, preview exactly what gets removed, and clean them up — free, in your browser.
Drop a watermarked PDF here
How watermark removal works
This tool reads the text on every page of your PDF and looks for two signature watermark patterns: text that shows up at the same position and rotation across most pages, or text that repeats many times in a rotated, tiled grid on a single page — the two most common ways a "CONFIDENTIAL" or "DRAFT" stamp gets added to a document. Rather than asking you to trust a text description, you get a live preview of an actual page for each match, with the exact area that will be covered highlighted, so you can confirm it found the right thing before removing anything. You choose which detected watermarks to remove, and each selected occurrence gets covered with a white rectangle wherever it appears — the rest of the page's content, layout, and formatting is left exactly as it was.
This is a text-layer tool, not an image editor or an AI content remover — it can't detect or erase a logo/image watermark, and it can't fix a watermark that's baked into a scanned page's pixels. For those cases, there's no separate text layer to find or remove, and doing it reliably would mean rebuilding the page's image content rather than editing text — a different tool entirely.
Common use cases
- Cleaning up an internal draft stamp before sending the final version to a client
- Removing a trial/preview watermark from a document you generated yourself with another tool
- Clearing a tiled "CONFIDENTIAL" stamp covering a single-page document like a receipt or form
- Preparing a clean copy of your own report for submission or archiving
- Stripping a repeated page-numbering or confidentiality stamp added by a previous export step
Because detection and removal both run entirely on your device, there's no upload wait, and your document never leaves your browser.
Related tools
Remove Watermark — frequently asked questions
Text that repeats at the same position and angle across most pages of the PDF, or the same rotated text repeated many times in a tiled pattern on a single page — the classic diagonal "CONFIDENTIAL" or "DRAFT" stamp, either way. It covers each detected occurrence with a white box and leaves everything else on the page untouched.
Each detected watermark is shown with a live preview of an actual page it appears on, with the exact region that will be covered highlighted — not just a text description. You can uncheck anything that isn't really the watermark before removing.
No — this only scans and removes text-layer watermarks. A logo or image stamped onto every page isn't detected or removed by this tool; that's a fundamentally different (and much harder) detection problem.
Yes, if the watermark is tiled — repeated many times at a rotated angle across that one page, which is the most common style for a single-page stamp. A watermark that only appears once, isn't rotated, or is very sparse on the page may be too ambiguous to detect automatically.
No — if the watermark is baked into a scanned or rasterized page, there's no separate text layer to find, so this tool can't detect or remove it.
No. This tool is meant for removing watermarks from your own documents — for example, a draft stamp you added yourself, or one from a tool you used that you have the right to remove. Don't use it to strip copyright or ownership marks from files that aren't yours.
It's always white — the tool doesn't sample your page's actual background. On a plain white page this is invisible; on a colored or textured page, the cover box may be visible. The preview shows you exactly where it will sit before you commit.
Yes — a Cancel button is available while pages are being scanned, so you're not stuck waiting through a huge file if you picked the wrong one.
Every distinct repeated stamp is listed separately with its own preview and checkbox, so you can remove some and keep others — useful when a document has both a "DRAFT" stamp and a separate page-numbering watermark, for example.