How to add a watermark to a PDF

This is the most important thing to be clear about. A watermark is text drawn onto the page. It does not encrypt anything, does not restrict what anyone can do with the file, and does not prevent copying, printing or forwarding. Someone who receives a watermarked PDF has the whole document and can read every word of it.

It can be removed, too. The text is page content, so an editor can strip it, a determined person can crop banners away, and rasterising and re-OCRing the page reproduces the content without the stamp. Tiled placement makes removal more tedious, not impossible.

Which is fine, because that is not the job. The job is communication: a reader who opens a page stamped DRAFT will not mistake it for final, and a recipient of a page marked CONFIDENTIAL knows the expectation before deciding to forward it. Marking a document sets expectations and creates a record of what was communicated. If you need to control access, encrypt the file — and understand that these are complementary measures, not alternatives.

Opacity is a balance between two failures. Too faint and the watermark is missed entirely, especially on a printout or a phone screen. Too strong and it competes with the text underneath, which is worse — a document people struggle to read is a document people ask you to resend without the stamp.

The default of 25 percent is a deliberate compromise and holds up for most documents. Around 15 percent suits dense text pages where legibility is paramount; 40 to 45 percent suits a classification marking that must be impossible to overlook, and works better as a banner than across body text.

Placement interacts with your document’s layout. Diagonal crosses everything, which is maximally visible and maximally intrusive, so it wants lower opacity. Top and bottom banners sit in the margin and barely interfere, so they can be much stronger — a bottom banner does need checking against existing page numbers. Tiling covers the most and needs the lowest opacity of all, since every part of the page has text over it.

A per-recipient watermark is the most useful variant: stamping a name or reference on each copy means a leaked document identifies the copy it came from. That is a deterrent and an audit aid rather than a technical control, and it is often more effective than encryption at changing behaviour.

Cover-page-only stamping is the lightest option — the pages field makes it a one-page selection — and it works when the concern is framing rather than page-by-page marking. Stamping every page is what you want when pages will be printed and circulated separately.

Two side effects worth knowing. The watermark text becomes part of the page text, so it shows up in searches and in text extraction, which occasionally confuses downstream processing. And on a scanned page there is nothing behind the stamp but an image, so the watermark sits over the picture of the text with no way for the reader to disambiguate — keep opacity lower on scans than you would on digital text.