Watermark PDF

The uncomfortable truth about drafts is that they travel. A contract revision goes to four people, one of them forwards it 'for context,' and eighteen months later an unmarked version of a deal you never signed surfaces in a dispute — indistinguishable, at a glance, from the real thing. Nothing on the page says DRAFT. Nothing ever did.

Same story, different office: HR sends a compensation summary to a manager for review, and it ends up photographed and passed around a break room. A photographer delivers proofs, and one turns up on a shop's wall. A consulting deck goes out under NDA, and three slides reappear in someone else's pitch, uncredited. In every case the fix is the same and almost embarrassingly cheap: put a mark on the page before it leaves your hands.

Watermark PDF does that marking in your browser — diagonal or horizontal text, any opacity, optional tiling across the entire page, or an image watermark if what needs to travel is your logo rather than a status word. No account, no upload, no per-document fee. The file opens locally, gets its ink, and saves back to your disk.

A watermark is a small design decision that people tend to get wrong in the same two directions: so faint it is decorative, or so heavy it drowns the document. The controls here exist to land between them.

• Text watermarks with free wording — DRAFT, CONFIDENTIAL, INTERNAL, DO NOT DISTRIBUTE, or anything your situation demands.
• Image watermarks: your logo or signature mark embedded once and drawn on every page.
• Rotation: the classic 45-degree diagonal, horizontal bands, or anything in between.
• Opacity control from whisper-faint to unmistakable, with a live preview so you can see the effect at real zoom.
• Tiling that repeats the mark in a grid across the page, so cropping a corner does not defeat it.
• Scale and margin controls, plus per-page choice — one marking style for the whole file, or a heavier hand on the sensitive pages.
• Both underlay and overlay placement: the mark behind the text like stationery, or on top of everything like a stamp.

Two design defaults carry most of the value. Diagonal reads as a status mark and resists the eye's habit of ignoring footers. Tiling survives cropping. Everything else is taste — and taste is what the preview is for.

The routine below is the common case: a confidential contract on its way to a counterparty. The same five moves apply to proof galleries, internal decks, and anything else that needs to wear its status on its sleeve.

1. 1

   Open the document

   Drag the PDF in. It renders locally, page by page, and stays on your machine for the entire operation.

2. 2

   Enter your text or image

   Type the watermark word, or load the logo PNG. Set the size relative to the page so it reads at print scale, not just at screen scale.

3. 3

   Set angle, opacity, and tiling

   Diagonal for status marks, tiled if leakage is a live worry. Start near 12 percent opacity and adjust while looking at your densest page, not the title page.

4. 4

   Preview across pages

   Flip to a text-heavy page, a table page, and a chart page. Watermarks interact with content differently on each — a mark that flatters a cover page can flatten a spreadsheet.

5. 5

   Save the marked copy

   Download the result. Your original stays unmarked on disk, which matters: the clean version is the one you eventually send when the draft becomes final.

Every PDF page is described by a content stream — a little drawing program of operators. Watermarking appends operators to that program. For a text mark, the tool sets a font, applies a rotation matrix for the diagonal, and draws the glyphs. For an image mark, it embeds the image once in the file as a shared object and references it from every page — which is why a logo on 200 pages costs the file size of one logo, not 200.

The text-versus-image distinction also plays out in file geometry. A text watermark is a handful of drawing operators and a reference to a font — its cost to the file is measured in bytes, which is why stamping a 500-page production set adds almost nothing to the transfer size. An image watermark embeds the image once, as a shared object, and every page's content stream references that same object. Pages do not each carry their own copy; they point at one. The format was designed economically, and the watermark rides those economics.

Opacity is handled by a structure the PDF standard calls an extended graphics state. It is a named set of drawing parameters — a transparency level among them — that the content stream applies before painting the watermark. This is the same machinery professional prepress tools use, and it is why a 12-percent watermark genuinely blends with the page instead of sitting on it like a gray rectangle.

The library underneath is pdf-lib, running in your browser. It parses your document, computes each page's dimensions from its media box, positions the mark per your settings — centered, tiled on a grid, or anchored to a corner — and writes the new content stream operators. Rotation is a transformation matrix; tiling is repeated draws at computed intervals; nothing about either requires a server.

Underlay versus overlay is worth thirty seconds of thought. An overlay draws after the existing content, on top of everything — the stamp model. An underlay draws before it, so the document's own text and images print over the mark, the letterhead model. Text documents usually want underlays so words stay readable; scans and photographs need overlays, because an underlay would vanish beneath the page image entirely.

One more technical note for the curious: because the watermark is vector ink drawn by operators — not a rasterized layer baked over a screenshot — your document's text stays selectable, searchable, and sharp. Search engines, screen readers, and copy-paste all continue to work. The watermark is a good citizen of the format.

Concrete cases make the options make sense. These four cover most of the watermarking world in miniature.

Honesty time: a watermark is a fence made of paint. It changes behavior and preserves evidence, but it does not stop a determined actor. Knowing exactly where the protection ends is what lets you use the tool well.

• It is not encryption. Opening control, permission flags, and certificate-based protection are separate mechanisms. If a document must not be readable by strangers, password-protect it — marking it instead is like locking a door with a poster.
• It can be cropped or patched. Tiling raises the cost of cropping considerably, and any removal leaves editing evidence, but a committed editor with professional tools can obscure a mark. Watermarks raise the bar; they are not the ceiling.
• It does not track copies. There is no invisible serial number, no leak-tracing per recipient — those exist in dedicated services, usually by varying the mark per copy. If leak tracing matters, that is a different product category.
• It does not anonymize. A watermark adds identity to a document; it never removes it. Sensitive content stays sensitive, just better labeled.
• OCR noise. On heavily tiled, heavily opaque scans, text recognition quality can dip slightly because the mark crosses the glyphs. If a scanned document is headed for OCR, favor a single centered mark over a dense tile.

There is also a practical ceiling on how much marking a document can carry before it becomes unreadable, and it arrives faster than people expect. Tiling plus high opacity plus a wordy watermark — three reasonable choices — can combine into a page where the reader cannot find the actual text. The failure is gradual, each step defensible, the sum a mess. When in doubt, shorten the wording before raising the opacity. CONFIDENTIAL carries further than CONFIDENTIAL — DO NOT DISTRIBUTE — INTERNAL USE ONLY, at every setting.

Notice the pattern in those limits: every one of them has a real answer that is a different tool or a different workflow. Marking is one layer of document hygiene. It pairs naturally with signing the final version, with page numbering for citability, and with merging for assembly — none of which replaces the others.

Law firms are the heavyweight users — privilege and confidentiality markings are close to a professional obligation, and litigation drafts carry DRAFT stamps as a matter of hygiene. HR departments mark compensation files, org charts ahead of reorganizations, and investigation notes. Finance teams mark board materials, pre-release earnings summaries, and anything forward-looking that could move a stock if it wandered.

The creative trades run on image watermarks. Photographers mark proofs so that the review copies cannot quietly become the finals; designers mark portfolio PDFs that circulate for job hunts and client pitches; stock agencies have marked previews since the industry existed. The economics are blunt: an unmarked proof is a free deliverable, and a marked one is an advertisement at worst.

Government and healthcare round it out — FOIA response drafts, internal policy reviews, and the general administrative instinct that documents should declare their sensitivity themselves. The shared motive across every field: pages outlive their contexts. A mark travels with the page into contexts you cannot control, saying what you would say if you were standing there.

A watermark serves two readers at once, and they want opposite things. The intended recipient needs the document readable; the future stranger — the person who encounters a leaked copy — needs the mark unmissable. Opacity is the negotiation between them, and there is no single correct setting, only a correct setting per document.

The variable that matters most is the density of the underlying page. A contract set in ten-point text with tight leading drowns quickly; ten to twelve percent opacity keeps the mark legible without taxing the reader's eyes. A slide deck with generous whitespace can carry twenty percent without complaint. Gray is the default because it recedes; red earns attention for genuinely urgent restrictions but tires quickly across a hundred pages. Reserve it for the documents where the mark itself is doing legal work.

One accessibility note worth internalizing: text watermarks live on a separate layer from the document text, so screen readers continue to read the body unimpaired — but heavy tiling over dense text does affect low-vision readers and OCR accuracy, because it lowers the effective contrast of the glyphs beneath. If a document is headed for scanning-back or machine reading, favor a single centered mark over a dense grid. The mark should inconvenience the misuser, not the legitimate reader.

Before a file leaves your machine wearing a mark, run it through this sequence. It is short enough to actually get used, which is the whole point of a checklist.

• Name the worst realistic day for this document — leaked, forwarded, photographed, cropped — and let that day choose the style: single mark or tile.
• Decide what the mark says in the recipient's language: DRAFT for versions, CONFIDENTIAL for sensitivity, INTERNAL for circulation limits, your name or logo for authorship.
• Set the angle: diagonal for status marks, horizontal for brand marks that must stay upright in cropped screenshots.
• Dial opacity against your densest page, not your cleanest one.
• Flip through a text page, a table page, and a full-bleed image page before committing.
• Save the marked copy under a new name, and archive the untouched original somewhere you will find it in a year.

The pass takes about five minutes on a typical document and repays itself the first time a marked page is discovered somewhere it should never have been — the moment when the mark, not your memory, is what speaks.

A rule that has aged well: mark for your worst realistic day, not your worst imaginable one. A negotiation draft needs one honest diagonal. A file that would genuinely hurt if photographed in a break room earns the tile. The image options buy identity; the text options buy status. Both buy you the thing a watermark really sells, which is the ability to say, later, with evidence: that page was marked, and someone chose to ignore it.

There is an irony at the center of most watermarking services: to mark your confidential document, you must first upload it — the whole file, sensitive contents included — to their servers, and trust their storage, their retention policy, and their breach posture. The mark says CONFIDENTIAL on a page that already left the building.

This tool never makes that trade. pdf-lib executes in your browser, the watermarking is local computation, and the marked file is written straight back to disk. There is no upload step to inspect in the network tab because there is no upload. Pull your network connection after the page loads and mark a 300-page production set on a train through a tunnel.

For the documents that most often get marked — legal, medical, financial, unreleased creative work — we would argue this is not a feature but the feature. The privacy model and the purpose of the tool point the same direction. The rest of the PDF suite here is built on the same principle, from PDF Merge to splitting and rotating: your files compute on your silicon, under your roof.

Everything above compresses into the questions below. If your question is not here, the honest short answer is usually: preview it, and keep the unmarked original.

Watermarks rarely work alone in a professional workflow. Sign PDF closes the loop when the draft becomes final — the unmarked original gets a signature instead of a stamp. Add Page Numbers to PDF makes marked documents citable page by page, which matters when the marked file is a draft deposition or a numbered exhibit. And PDF Merge assembles the pieces before any of them wear ink. Same suite, same rule: everything runs locally, and none of it asks for an account.