What a watermark is for
A watermark makes a claim about a document's status, visible to anyone who opens it. This is a draft. This is confidential. This copy belongs to a particular person. It works by being unavoidable: the reader cannot look at the page without also seeing the claim.
That is the entire mechanism, and it is worth being clear about it, because a watermark is routinely reached for as though it were a security control. It is not one. It is a label, and labels change behaviour by informing people rather than by preventing them.
It does not protect anything
Anyone with a PDF editor can remove a text watermark in about a minute. It is drawn into the page as an object like any other, and deleting an object is what editors are for. Even without editing, a determined person can screenshot each page, or select and copy the underlying text, which is untouched.
So the honest description is deterrence plus attribution. A CONFIDENTIAL stamp does not stop a leak; it removes the excuse that nobody knew, and it makes forwarding feel like a decision rather than a reflex. A per-recipient stamp does not stop copying; it means a leaked copy points at whoever it was issued to, which is a real effect on behaviour precisely because everyone understands it.
If a document genuinely must not be read by the wrong person, the answers are elsewhere: a password on the file, permissions that disallow copying and printing, or simply not sending it. Those have their own weaknesses — PDF permissions are widely ignored by readers that choose to — but they are at least attempting the job a watermark is not attempting.
Redaction is a different thing entirely
Worth stating because the confusion is expensive. Drawing a black box over a name is not redaction. The text is still in the file underneath, and selecting the area copies it out. This has embarrassed courts, newspapers and government departments repeatedly.
Real redaction removes the underlying content and leaves the mark. It needs a tool built for it, and it needs verifying afterwards by searching the file for the text you thought you removed. A watermark, a black rectangle and a redaction look similar on screen and are three unrelated operations.
Making one that reads without ruining the page
The whole difficulty is that a watermark has two jobs in tension: be noticed, and do not obstruct. A few things settle it in practice.
- Opacity around fifteen to twenty per cent. Dark enough to register, light enough that body text stays comfortable and the page still prints legibly. Above about thirty the document becomes tiring to read; below about ten people genuinely miss it.
- Diagonal rather than horizontal. A diagonal stamp crosses more of the page, is harder to crop out, and reads as a stamp rather than as content. Horizontal placement can be mistaken for part of the document.
- Short text. One or two words. DRAFT, CONFIDENTIAL, a date, a name. A sentence set large enough to see becomes an obstruction, and set small enough to ignore becomes pointless.
- A colour that survives greyscale. Most documents get printed on an office machine in black and white. A pale red at low opacity can vanish entirely; check what it looks like without colour before committing.
The [msm_tool slug="pdf-watermark"] draws the stamp into each page as Helvetica text, sized against that page's own diagonal, which is why a single setting works across A4, Letter and a landscape slide without adjustment.
Text stays text
A watermark added properly does not flatten your document. The words underneath stay selectable, searchable and accessible to a screen reader, and the file stays a real PDF rather than becoming a stack of images.
That distinction matters more than it sounds. The common alternative — printing to an image and stamping that — destroys the text layer, multiplies the file size, and makes the document unsearchable and unreadable to assistive software. It is a much larger cost than the watermark is worth.
Which pages to mark
Usually all of them, because pages get separated. A single page printed from the middle of a document, or a screenshot of one section, travels without its context, and marking only the cover means the rest carries no claim at all.
The exceptions are deliberate. A cover page carrying a formal notice may not want the stamp repeated over it. A document whose first page is a title card sometimes reads better with the stamp starting at page two. Both are editorial calls rather than defaults.
Where it fits in a workflow
Mark at the moment of sending rather than at the moment of writing. A DRAFT stamp baked into the working file has to be removed before the final version goes out, and the version where somebody forgot is the one that reaches the client.
Keep the unmarked master, stamp a copy for each recipient or each stage, and treat the marked file as disposable. That way the stamp always says something true about the copy in front of the reader, which is the only property that makes a watermark worth adding at all.