Image watermarks
Pictures repeated across pages at the same position — logos, seals, scanned signatures — identified as one candidate with every page it appears on.
See the difference
Drag the handle. Same page, same text, same table — only the watermark goes. Pick a mark type to see each of the four the detector removes.
⋮
Which watermarks can actually be removed
Removability is not a matter of how good the tool is. It is decided entirely by how the watermark was stored in the file — and you can know which case you have before you upload anything.
A watermark that is its own object — a layer, an annotation, an image, a form overlay — can be deleted and the page underneath is untouched, because the two were never joined. A watermark that has been flattened shares its drawing instructions with the content beneath it. Removing it removes that content too.No tool gets past this. It is a property of the file, not a limitation of the software.
| Type of mark | How it is stored | Outcome |
|---|---|---|
| Layer watermark | Attached to an optional content group — a PDF layer, resolved at render time rather than merged into the page. | Removes cleanly |
| Tool-added text watermark | Added through an application’s watermark feature, which stores it as its own recognisable layer object. | Removes cleanly |
| Annotation stamp | A stamp annotation such as DRAFT or APPROVED, which is a separate object class from a watermark. | Removes cleanly |
| Image or logo overlay | A picture or form object placed on each page, independent of the text around it. | Removes cleanly |
| Plain page text | Not a watermark object at all — just text written into the page content stream, which is why watermark features ignore it. | Usually removable |
| Mark over the content | Sits on top rather than behind. The object still comes out, but anything it visually replaced was never there to restore. | Depends on overlap |
| Flattened into the page | Merged with the page content, so the mark and the text beneath it are the same drawing instructions. | Not cleanly removable |
| Burned into a scan | Part of a single raster image. There is no separate object, only pixels shared with the content. | Not removable without damage |
If your file falls in the bottom two rows, the honest answer is to ask whoever produced it for a copy without the mark. Covering it and re-flattening leaves a worse document than the one you started with.
Why “Remove Watermark” often does nothing
The most common reason people end up looking for another tool, and it is not a bug.
Acrobat’s Remove Watermark removes watermarks Acrobat added. It writes them as a recognisable layer object, and the feature looks for exactly that object. A mark that arrived from Word, from a scanner, from a report generator, or drawn straight into the page content is not that object — so the menu item runs, reports success, and changes nothing. The same applies to stamps, which are annotations rather than watermarks and live somewhere else entirely.
This detector works the other way round: it looks for repeated elements at consistent positions across pages, scores them, and reports what it found regardless of which application put them there. That is why it finds marks a watermark feature will not, and also why it shows you its reasoning instead of a single button.
How to remove a watermark from a PDF
Three steps, and nothing is changed until you say so.
Upload your PDF
Drag and drop the watermarked PDF, or choose a file. Every page is scanned for watermark objects.
Review and confirm
See what was detected — image, text, layer, or stamp watermarks — and tick the ones to remove. Nothing changes until you confirm.
Download the clean PDF
Remove the confirmed watermarks and download a clean file — the rest of the page stays exactly as it was.
Scope is yours: every page, only the page you are looking at, or a custom range such as 1-3,7. A mark repeated through the document is detected once, with all of its pages listed.
What the detector reports
Six things, including the ones that mean “this will not work”.
Text marks
Both proper text watermarks and text simply drawn into the page content, which is the case an application’s own Remove Watermark button will not touch.
Layers
Optional content groups carrying the mark. These are the cleanest case: the content underneath was never merged with the watermark.
Annotation stamps
DRAFT, CONFIDENTIAL, APPROVED and their relatives, stored as annotations rather than page content — a different object class that people routinely confuse with watermarks.
Confidence and signals
Each candidate is classified as a likely watermark, a possible one, or reported for transparency — with the signals behind the score, so you are not asked to trust a verdict blind.
The honest flags
Flattened pages, scanned pages, sampled coverage on long documents, and a signed document are all reported up front rather than discovered after a failed removal.
It checks its own work
Removing an object is easy. Confirming you removed only that object is the part worth doing.
Safety check before applying
Each candidate carries a removable flag and any safety notes attached to it, so a removal that would take content with it is flagged rather than attempted.
Re-rendered and compared
After removal the pages are rendered again and compared against the original, so collateral change is caught here rather than by you, later, in the file you already sent.
Signed files flagged first
A digital signature covers the content, so any edit invalidates it. Signed documents are detected and called out before anything is applied — remove first, sign after.
Only on documents that are yours to change
Whether removing a watermark is fine or a problem has nothing to do with the technique. It depends on whose document it is.
What this is for
Your own files and files you are authorised to modify: an outdated logo on a template you maintain, a DRAFT or COPY stamp on a document now final, a trial mark on output you generated, internal branding being retired. Ordinary document work.
What it is not for
Someone else’s copyright, ownership or attribution mark, and stock, library or vendor watermarks on work that is not yours. In the US, removing copyright management information carries statutory damages in its own right, separate from any question of infringement. Ask the owner for a clean copy instead.
Removing a mark does not remove the obligations that came with the document — confidentiality, licensing and copyright all still apply. On engineering and architectural drawings, a superseded or approval stamp is part of the record: issue a new, correctly stamped revision through your document-control process rather than editing the old one. This is general information, not legal advice.
When it will not come off
Six symptoms, what each one actually means, and whether anything can be done.
| Symptom | What it means | What to do |
|---|---|---|
| Acrobat’s Remove Watermark does nothing | It only removes marks Acrobat itself added and stored as a watermark layer. | Anything from Word, a scanner or another application needs object-level detection instead. That is what this tool does. |
| The watermark cannot be selected | It is on a layer, in an annotation, or flattened into the page — none of which behave like selectable text. | Upload the file and read which type it is. Selection is not how object-level marks are removed. |
| A white box is left behind | The mark was covered rather than deleted, so a filled rectangle now sits where it was. | Removing the object leaves no box. If a cover is the only option, the mark was flattened. |
| It is on every page and removal missed some | The mark is stored per page and only the current page was in scope. | Set the scope to all pages, or give a custom range such as 1-3,7. |
| The signature is now invalid | Any content change alters the document hash a signature covers. | Unavoidable — remove the mark first, then sign. Signed files are flagged before anything is applied. |
| Nothing was detected | The mark is flattened, part of a scan, or genuinely not a repeated overlay. | The report says which. A flattened mark needs the original file, not a better remover. |
Limits, and what happens to your file
Stated up front, because finding out afterwards is worse.
No OCR and no inpainting
Nothing is reconstructed. Content that was flattened under a mark is gone from the file, and this tool will tell you that rather than paint over the gap and call it removal.
One file, honest coverage
Up to 10MB and 100 pages per run. On long documents detection samples pages and says so, and encrypted files must be unlocked first.
Processed, then removed
Your file is sent over an encrypted connection, processed, and removed afterwards. It is not stored, shared, indexed or used to train models, and nothing is added to your output.
Detail in file handling and the privacy policy. Want to see what a mark is made of first? The object explorer shows the layers, annotations and image objects directly, and the layout validator confirms nothing else moved.
PDF Watermark Remover FAQ
Upload the file and the tool scans every page for watermark objects — image, overlay, layer, annotation and text marks. You review what it found, tick the ones you want gone, and download a clean file. Nothing is changed until you confirm.
No, and any tool claiming otherwise is overselling. A watermark stored as its own object comes out cleanly. One that has been flattened or burned into a scan shares pixels with the content underneath it, so removing it would take that content with it. The report tells you which kind you have.
Because flattening merges the mark into the page content, so the watermark and the text beneath it are no longer separate things. Deleting it deletes whatever it overlapped. The only options left are covering it or recreating the page, both of which lose fidelity.
Not cleanly. A scanned page is one image, and the watermark is part of that image, so there is no object to delete. The tool detects scanned pages and reports them as such rather than pretending to succeed.
Because it only removes watermarks Acrobat itself added, which it stores as a recognisable layer object. A mark added by Word, by a scanner, by another application, or drawn as ordinary page text is invisible to that feature — which is the single most common reason people end up looking for another tool.
Not for object-level marks. Removing an object leaves everything else untouched, because the rest of the page was never part of it. After removal the tool re-renders the pages and compares them against the original, so unexpected collateral change is caught rather than shipped.
No. A white box is what you get from tools that cover the mark instead of deleting it. Removing the object itself leaves the page as it was before the watermark was added — including whatever content sits behind it.
Not when the mark is a removable object — nothing is re-encoded and the page is not rasterised. Quality loss comes from the other approaches: converting pages to images, or covering and re-flattening the document.
Yes. A watermark repeated across the document is detected as one candidate with all its pages listed, and you can apply removal to every page, only the page you are viewing, or a custom range such as 1-3,7.
Yes. Any change to the content changes the document hash, which invalidates a signature that covers it. Signed files are detected and flagged before anything is applied — remove the mark first and sign afterwards, in that order.
No. The output carries no branding, logo or mark of any kind, there is no signup, and the file is not watermarked as a trial. Some free tools do exactly that; this one does not.
Because hiding its reasoning would be worse. Candidates are classified — likely watermark, possible watermark, or simply reported for transparency — with the signals behind each score. You decide what to remove, so seeing a rejected candidate is more useful than not knowing it was considered.
Image watermarks placed as a picture object, overlay form objects, optional-content layers, annotation stamps such as DRAFT or APPROVED, and text drawn straight into the page content. Each is stored differently, which is exactly what determines whether it can be removed.
A PDF layer. Content attached to one is suppressed at render time rather than deleted, which is why layer watermarks come away cleanly — the material underneath was never merged with the mark and is restored in full.
Yes, when it is stored as its own object — logos are usually image objects and DRAFT or CONFIDENTIAL marks are often annotation stamps or layers. Both show up as candidates alongside conventional watermarks.
It depends entirely on whose document it is. Removing a mark from your own file, or one you are authorised to modify, is ordinary document work. Removing someone else’s copyright, ownership, attribution or stock-library mark can breach copyright law, and in the US removing copyright management information carries statutory damages of its own. This is general information, not legal advice.
When the watermark is not yours to remove — stock images, licensed libraries, vendor or publisher marks, or anyone else’s attribution. Ask the owner for a clean copy instead. Removing the mark also does not remove confidentiality, licensing or copyright obligations that came with the document.
Do not quietly strip it. On engineering and architectural documents a superseded or approval stamp is part of the record, and the correct step is issuing a new, properly stamped revision through your document-control process rather than editing the old one.
The PDF is sent over an encrypted connection, processed, and removed afterwards. It is not stored, shared, indexed or used to train models, and no account is required.
Find out what kind of watermark you have
Free, no signup, and no watermark added to your output. You see what was detected before anything is removed.
Remove a watermarkAll PDF toolsEdit PDFPDF object explorerLayout validatorCompress PDF