Free PDF Tools That Work in Your Browser

Which tool solves which problem

Almost nobody wakes up wanting to "use a PDF tool". They have a document that is wrong in some specific way and they want it to stop being wrong. The table below is organised the way the problem actually arrives — as a complaint, not as a feature name.

Several of these chain together in one sitting. Merging a set of scans, cropping the grey scanner edges off every page, running OCR so the result is searchable, and then adding page numbers is four operations — but it is one workspace and one export. You do not download an intermediate file between each step.

How browser-based PDF editing actually works

The phrase "in your browser" is used loosely across the web, and often means "uploaded to a server, then shown to you in a browser". Here it is meant literally: the code that opens, modifies, and writes your PDF is JavaScript and WebAssembly executing inside the tab you already have open.

Three distinct jobs happen, and it is worth separating them because they explain both the capabilities and the limits described further down this page.

Reading and displaying a page

A PDF is not a picture of a document — it is a set of drawing instructions plus embedded fonts and images. To show you page four, the browser has to interpret those instructions and paint the result onto a canvas. That rendering step is what lets you see thumbnails, pick pages, and position a signature exactly where you want it rather than guessing at coordinates.

Writing a new file

Every operation on this site produces a brand-new document. Reordering pages rewrites the internal list that records page order. Adding a signature appends fresh drawing instructions to a page's content stream. Merging builds a new file whose page list points at content copied from each source. Your original file is opened, read, and left exactly as it was — the export is a separate object that never overwrites it.

Recognising text in a scan

OCR is the heaviest of the three, which is why it historically required a server. A recognition engine compiled to WebAssembly changes that: the model runs at close to native speed inside the tab. It examines the shapes on each scanned page, decides which characters they represent, and writes an invisible text layer positioned over the image. The page still looks like a scan, but selection and search now work.

The practical consequence of all three running locally is simple. There is no upload step to wait for, no queue behind other users, and no copy of your document sitting on a machine you do not control. Read more abouthow browser-based processing protects privacy.

Browser processing compared with uploading

Both models can produce a correct PDF. They differ in what has to happen to your document along the way, and those differences are the reason this site exists.

The retention row is the one that matters most for anything confidential. A privacy policy is a promise about what a company intends to do with a file it already possesses. Local processing removes the premise: there is no copy on anyone else's disk, so retention, breach, and deletion requests are all moot.

What browser-based tools can and cannot do

Being straight about the limits is more useful than a feature list, because most frustration with PDF software comes from expecting a capability that was never there.

Editing existing text is not one of these tools

This is the single most common misunderstanding, so it is worth stating plainly. Add Text places a new layer of characters on top of a page. It does not reach into the original content stream and change the words that are already there. If you need to correct a typo inside an existing paragraph, the reliable route is to fix the source document and export a fresh PDF. Anything that claims to rewrite arbitrary PDF text is usually reconstructing the page and quietly changing its layout, fonts, or spacing.

Very large files are bounded by memory, not by policy

There is no upload cap here because there is no upload. What there is instead is your device: the document is held in the browser tab while you work on it, so a several-hundred-megabyte scan of a long report will strain a phone before it strains a laptop. If a large file struggles, closing other tabs frees real memory, and splitting the document into parts is usually faster than fighting it as one object.

Protection is encryption, not a permission promise

Setting a password to open a document encrypts its contents, and that is genuine protection. Permission flags — "do not allow printing", "do not allow copying" — are a different thing: they are requests recorded in the file that well-behaved readers honour and others ignore. Treat the open password as security and the permission flags as a courtesy, and you will not be surprised later.

OCR accuracy depends on the scan, not the effort

Recognition works from the pixels it is given. A straight, well-lit 300 DPI scan of printed text produces excellent results. A phone photograph taken at an angle in poor light, or a page of handwriting, will not — no setting compensates for information that was never captured. Cropping away scanner edges and rotating pages upright before running OCR measurably improves the outcome.

Jobs that take more than one tool

The table above matches one complaint to one tool, which is the right answer for most of them. The jobs that go wrong are the other kind: the ones that need three or four operations in a particular order, and the ones where the obvious tool turns out to do nothing at all.

When the obvious tool is the wrong one

Five instincts come up again and again, and each one is reasonable. We measured all five rather than arguing about them, and four of the five do nothing useful at all.

The pattern behind four of those five is the same. Cropping, splitting, and zipping all change how a document is presented or divided without changing what it contains, and only operations that genuinely discard content — removing pages, re-encoding images — move the number. The methods behind every figure are onthe measurements page, andPDF too large to email works through the size problem end to end.

Problems no PDF tool can fix

• A blurry scan cannot be sharpened into detail the scanner never captured. Rescanning at 300 DPI is the only real fix.
• Two or three degrees of skew is not a quarter turn. Straightening that belongs in the scanner's own deskew setting.
• A forgotten password cannot be recovered. Unlocking requires the correct one, and there is no back door to offer.
• A file that will not open anywhere may not be a PDF at all, or may have been truncated by an interrupted download.
• Data inside embedded images — including GPS coordinates in phone photographs — is not touched by any page operation. What hidden data a PDF carries covers it.

Understanding the PDF you are holding

Two files can both end in .pdf and behave completely differently. Knowing which kind you have explains most of the surprises. None of it is a quirk of one program: page boxes, compression filters, and encryption are all defined byISO 32000-2, the international standard behind PDF, and the PDF Association's reference material on that standardexplains most of what follows.

Digital PDFs versus scanned PDFs

A digital PDF was generated by software — exported from a word processor, printed to PDF, produced by an invoicing system. Its text exists as characters, so selection, copying, and search all work immediately.

A scanned PDF came from a scanner, a copier, or a phone camera. Each page is a single image. There are no characters in it at all, which is why searching a scanned PDF finds nothingno matter how clearly you can read it yourself. OCR is what bridges the gap.

A third case trips people up: a hybrid where a digital cover page is followed by scanned attachments. Search works perfectly on page one and fails from page two onward, which reads like a broken search rather than a change in page type.

Two kinds of password

A user password is required to open the file at all — without it, nothing displays. An owner password leaves the file freely openable but marks certain actions as restricted. When a PDF opens instantly yet refuses to let you copy a line, an owner password is usually the reason.

Why file size varies so wildly

A forty-page text document can be under a megabyte, while a four-page scan can be forty. Characters and vector shapes are compact; photographs of paper are not. Scanning at 600 DPI in full colour when 300 DPI greyscale would do is the usual cause of an unexpectedly enormous file. Cropping the margins away does not help: it sets a smaller visible box and leaves the image data exactly where it was, so the file grows a little rather than shrinking — we measured it at 135 bytes larger.

Working through a document, in order

Most real jobs are not one operation but four or five, and the order they run in changes the result. A page recognised before it is straightened produces nonsense; a document signed before it is finished has to be signed again. Every figure below was measured on these tools rather than repeated from elsewhere — the same three-page, four-megabyte scan throughout, with the methods on the measurements page.

The crop follows the rotation

One step in that sequence deserves a paragraph of its own, because it used not to work this way. A crop is measured from the page as it is displayed — the orientation you are looking at in the preview, whether the file arrived turned or you turned it here. The margin labelled top trims the edge at the top of the preview.

Until we measured it, it did not. A page carrying a quarter turn is stored unturned with a rotation flag, and the crop used to be written against the stored page: typing 100 into top on such a page took 100 points off the edge showing on the right. It now trims the edge you were looking at, at every quarter turn, and the order you rotate and crop in no longer changes the result.

The step that resets your document's identity

Page operations divide into two kinds, and the split is invisible from the outside: some edit the pages already in your file, while others build a new document and copy the pages into it. Only the second kind loses anything, and what it loses is everything that belonged to the document rather than to a page.

Merging is the starkest case. Given two files with different titles and different authors, the merged result carried neither — both fields came back empty. It does not inherit from the first file, or the largest one, or the one you added first.

That cuts both ways, so read it in whichever direction applies. The author field is filled in automatically by most software from your computer account, frequently with a real name, and a rebuild strips it without being asked. Useful — but not a redaction guarantee, because it leaves data held inside embedded images untouched, whichwhat hidden data a PDF carries covers. And if the finished document needs a proper title and author, set them after the rebuilding steps rather than before.

Where the order does not matter

• Watermarks and page numbers are drawn in place and can go almost anywhere — after the page set is final, and before signing.
• Rotation can be repeated as often as you like. Nothing is decoded, so nothing accumulates damage.
• Searching can be run at any point, including on a file you have no intention of changing.
• Every tool here produces a new file and never modifies the one you opened, so a wrong sequence costs you the export, not the document.

Privacy, in specific terms

"We respect your privacy" is a sentence every site writes. What follows is what it actually means here, expressed as things that either happen or do not.

• Your PDF is never transmitted. There is no upload endpoint for document content to reach.
• Filenames and metadata are not sent anywhere either — they stay with the file, in the tab.
• No account exists, so there is no profile to attach activity to.
• Nothing is stored between sessions. Closing the tab ends it; there is no history to clear.
• Exports are assembled on your device and handed to your browser's own download mechanism.

This is a structural property, not a policy commitment — which is the important distinction. A promise not to look at your files depends on the promise holding. An architecture with no copy of your file cannot look at it regardless of intent. The privacy policy and the security pageset out the details, and the editorial policy covers how the guidance here is written and reviewed.

Why PDF Editing Tools

Most online PDF tools ask you to upload your document to a server before they'll touch it. PDF Editing Tools was built around a different premise: your PDF editing tool should not need to see your files at all. Every merge, signature, crop, and password is computed in your browser tab, using the same JavaScript engine that renders this page.

That means PDF Editing Tools works the same way whether your document is a public flyer or a private contract — the processing model doesn't change based on sensitivity, because nothing ever leaves your device in the first place.

PDF terms explained

Words that appear throughout this site and in the guides, defined once so the rest reads clearly. They are grouped the way the questions tend to arrive: what kind of file you are holding, what is inside its pages, how its text and fonts behave, what survives an edit, and what locking or signing it actually means.

What kind of file you have

Digital PDF

A PDF generated by software, where text is stored as characters. You can select, copy, and search it directly.

Scanned PDF

A PDF whose pages are photographs of paper. To a computer each page is one image, so there is no text to select until OCR creates some.

Hybrid PDF

A document that mixes both — a digital cover page followed by scanned attachments, for instance. Search works perfectly on the first page and finds nothing from the second, which reads as a broken search rather than a change of page type.

PDF/A

An archival profile of PDF that forbids anything whose meaning could drift over time, such as a font that is named but not carried in the file. A document can claim conformance in its metadata without actually conforming.

Inside the page

Page tree

The internal structure listing a PDF page order. Reordering pages rewrites this list rather than moving the page content itself.

Content stream

The drawing instructions for one page. Adding text writes new instructions on top; it does not rewrite existing ones.

MediaBox

The sheet — the full physical page as the file declares it. This is what page size means to a printer, a layout program, or a submission system that checks dimensions.

CropBox

The window — the region a viewer or printer actually shows. Cropping sets this box and leaves everything outside it in the file, which is why a crop adds bytes instead of saving them.

Page rotation

A quarter-turn instruction stored on the page rather than a redrawing of its content. The page keeps its declared dimensions: turned 90 degrees, a portrait page displays landscape but still reports portrait size.

Point

The unit a PDF measures in. One point is a seventy-second of an inch, so 36 points is half an inch and a US Letter page is 612 by 792.

Content layer

New material drawn on top of a page — text, a signature, a watermark, page numbers — rather than an edit of what was there already. It cannot reflow the paragraphs underneath it.

Text, fonts and recognition

OCR

Optical Character Recognition — reading the shapes in an image and deciding which characters they represent, producing a text layer over the picture.

Text layer

Invisible characters positioned over a scanned image so that selection and search work while the page still looks like the original scan.

Base-14 fonts

The fourteen typefaces every PDF viewer is required to provide, Helvetica among them. Text set in one needs no font carried in the document, which is why adding a line costs a few hundred bytes rather than several hundred kilobytes.

Embedded font

A font program carried inside the document so the letters render identically everywhere. The trade is size — embedding is what makes a text-only PDF larger than its word count suggests.

Font substitution

What happens when a font is named but not embedded: the viewer supplies its own, usually Arial in place of Helvetica on Windows. Spacing stays fixed because the character widths are stored in the file; only the letter shapes change.

DPI

Dots per inch — how finely a scanner samples the paper. 300 DPI in greyscale is enough for text, while 600 in colour costs around twelve times as much for no gain in readability.

Skew

The two or three degrees of tilt a sheet picks up going through a feeder. It is not a quarter turn, so rotation does not correct it; the deskew setting in your scanner software does.

What survives an edit

Rebuild

Copying pages into a fresh document, which is how merging, splitting, extracting, and reordering all work. Page content, size, and orientation come through faithfully; everything belonging to the document rather than to a page does not.

Document properties

The title, author, subject, and creator fields. Most software fills the author in automatically from your computer account, frequently with a real name. A rebuild clears them, while rotating, cropping, and watermarking leave them intact.

Bookmarks

The navigation pane, stored as a document outline rather than on any page. No rebuild carries it across, because nothing in the copying knows it existed.

Flattening

Merging separate layers — form fields, annotations, signatures — into the page itself so they can no longer be edited separately.

Object stream

A compression container holding many of the internal dictionaries of a file at once. It is the reason searching the raw bytes of a saved PDF for a value it definitely contains can come back empty.

Incremental save

Appending changes to the end of a file instead of rewriting it, so the earlier bytes stay exactly as they were. It is how a document records an edit without disturbing what a signature already covered.

Security and signing

User password

The password required to open a PDF at all. Without it the file cannot be displayed.

Owner password

A password that sets permissions — printing, copying, editing — on a file that still opens freely.

Encryption

Scrambling the streams inside a file so a password is needed to read them. It costs almost nothing in size: protecting a four-megabyte scan added around a thousand bytes.

Electronic signature

A mark showing intent to agree — drawn, typed, or uploaded as an image. It is page content, so it survives later edits, and it is what almost every everyday document is asking for.

Digital signature

A certificate-based cryptographic seal proving a file has not changed since it was signed. Any later edit invalidates it, including one as small as turning a page.

Form fields

Interactive boxes a reader can type into, held in a structure separate from the page content. A rebuild can lose or alter them, which is why a filled form is worth checking after any page operation.

Redaction

Permanently removing content rather than hiding it. Cropping a margin, drawing a black box, and covering text with a watermark are all reversible, so none of them is redaction.

Sending and running

Base64

The encoding email uses to carry attachments, turning every three bytes into four characters. That fixed one-third increase is why a stated 25 MB limit really carries about 18.8 MB.

WebAssembly

A compiled format browsers run at near-native speed. It is what makes local OCR practical without a server.

Delete, extract, or split: what happens to the original PDF?

Three of the most-used tools here accept the same input and the same page selection, and people reasonably assume they are variations of one operation. They are not. What separates them is the shape of what comes out, and choosing by output is faster than choosing by name.

Delete produces a single file: your document with the pages you chose taken out and everything else left in its original order. If you start with forty pages and remove four, you get one thirty-six-page PDF. Reach for it when the pages are unwanted rather than needed elsewhere — the walkthrough for removing pages you don't need covers the selection patterns in detail.

Extract also produces a single file, but the inverse one: a new PDF containing only the pages you chose, in the order you chose them. The same forty-page document with four pages selected gives you a four-page PDF. The pages you did not select are simply absent from the output rather than deleted from anything. If you need both halves, run the operation twice, or readextracting selected pages first.

Split is the only one of the three that produces more than one file. Instead of a selection it takes a rule — split at a page, split every N pages, split into individual pages — and applies it across the whole document, giving you several outputs at once. The guide tosplitting a PDF into separate files works through which rule suits which document.

One behaviour is shared by all three, and by every other tool on this site: the file already on your disk is never modified. Each tool builds a new document and hands it to you as a download, so an unwanted result costs you nothing but the export. That is also why the reverse operation is always available afterwards — merging files without uploading them will put pieces back together whenever a split or extract went further than you wanted.

What actually changes inside the file when you edit a PDF

A PDF is not a picture of a document. It is a container holding objects — page definitions, fonts, embedded images, an outline, metadata — with an index that records where each object sits. Understanding which of those an operation touches explains most of the surprises people run into after editing.

Some operations only change the index. Reordering, rotating and deleting pages rewrite the page tree, which is the list describing what order the pages come in and how each is oriented. The page content itself is untouched, which is why these operations are fast even on large documents and why they never degrade quality: nothing is re-encoded, only re-listed.Rotating pages is the clearest case of all, because the instruction it writes is a single value recorded against each page.

Other operations add new objects without removing anything. Adding text, a watermark or page numbers appends drawing instructions on top of a page's existing content. The original content stays exactly where it was underneath, which is why added text can usually be distinguished from the document's own text, and why these operations make a file slightly larger rather than smaller.

Cropping belongs to a third category that catches people out. It changes the box that defines the visible region of a page, not the content inside it. The material outside the new boundary stops being displayed but is still present in the file — which is why cropping does not reliably shrink a document and why it should never be used to hide sensitive material. The specifics are incropping margins without deleting content.

Knowing which category an operation falls into predicts the outcome better than any general rule about editing. It explains why a rotated page can revert in one viewer and not another, why a file grows after you annotate it, and why a document that looks trimmed still carries everything it had before.

A fourth distinction cuts across those three, and it decides what survives rather than how quickly the operation runs: whether the file is edited in place or rebuilt. Rotating a page and adding page numbers write onto the document that is already there. Merging, splitting, extracting and reordering build a new document and copy the pages into it — and a rebuild starts from an empty file, so it has nothing to inherit. Title, author and subject come back blank, any outline is gone, and a digital signature no longer validates, because a rebuilt document genuinely is not the document that was signed.

Text you can select, and text that only looks like text

Two documents can look identical on screen and behave completely differently the moment you try to search them. The difference is whether the words exist in the file as text, or only as the shapes of text inside an image.

A PDF produced by exporting from a word processor, a browser or most design software carries real text objects: characters, positions and font references. You can select a line, copy it, and search for a phrase, because the words are genuinely in the file. A PDF produced by a scanner or a phone camera usually carries a photograph of a page instead. Every word is visible and none of it is text — there is nothing to select because, as far as the file is concerned, the page is one picture.

This single distinction explains a whole family of frustrations: search finding nothing in a document you can plainly read, copy producing empty clipboard content, and screen readers announcing nothing at all.Why text won't copy works through both causes, because there is a second one — a document can contain real text that has simply been restricted.

Optical character recognition is the bridge between the two states. It examines the image, recognises the shapes as characters, and writes a text layer into the file positioned over the picture. The page still looks the same, but the words underneath are now real and can be searched, selected and read aloud.How OCR makes a scan searchable describes what that layer contains and where recognition tends to struggle.

Recognition is a reading of the page, not a transcription guaranteed to be perfect. Faint print, unusual typefaces, handwriting and skewed scans all reduce accuracy, so a searchable scan is best treated as a document you can now find things in rather than one whose every character is certain.

Telling the two states apart takes a few seconds and is worth doing before you plan any work around a document. Try to select a single word. If a text cursor appears and just that word highlights, the text is real. If the entire page highlights as one block, or nothing highlights at all, you are looking at a picture. Searching for a word you can plainly see on the page is the same test approached from the other direction, and it fails just as clearly.

Mixed documents are more common than people expect, and they defeat a single check. A report assembled from several sources can carry exported pages with real text alongside scanned pages without any, so a test on page one tells you nothing about page forty. Where the answer matters, sample a page from each part rather than whichever page opens first.

A document that has already been through recognition is a third state again. It has a text layer, so it searches and selects and behaves like a normal document — but that layer is a machine's reading of a picture rather than the document's own text, and it can be confidently wrong. A misread name or figure produces no error and no warning; it simply does not come back when someone searches for it later.

What a signature in a PDF actually is

Signing a PDF can mean two quite different things, and the word covers both. Knowing which one a situation calls for saves a great deal of confusion when a document is rejected or a signature is questioned.

The first is a visible mark: an image of a signature, or one drawn with a mouse or finger, placed onto the page. It is a picture positioned over the content, and it carries no information about who applied it or whether the document changed afterwards. For the very large number of everyday documents that simply need to show assent — an internal approval, a delivery note, a form returned to a colleague — it is exactly what is being asked for. Placing one is covered insigning a PDF online.

The second is cryptographic. A digital signature binds a certificate to the document's contents, so that altering so much as one character afterwards invalidates it. It answers questions the visible mark cannot: who signed, and whether anything has changed since. It also requires a certificate from a recognised authority, which is why it is not something a browser tool can produce on its own.

The distinction matters most when a recipient specifies which they require. A contract or filing that asks for a digital signature is asking for the cryptographic kind, and a drawn image will not satisfy it no matter how it looks.Electronic versus digital signatures sets out where each is normally expected and what verification looks like in practice.

What a PDF password does and does not do

PDF protection is narrower than it sounds, and the gap between what people expect and what the format provides is worth stating plainly before you rely on it.

There are two kinds of password. An open password encrypts the document: without it the file cannot be read at all, and its contents are genuinely unavailable. A permissions password leaves the document readable by anyone but marks certain actions — printing, copying, editing — as restricted. Both are set the same way, and password-protecting a PDF covers when each is the right choice.

The important difference is in how firmly each holds. Encryption from an open password is real protection: the bytes are scrambled and the password is required to unscramble them. Permissions are closer to a request. They are recorded as flags in the file and respected by well-behaved viewers, but the underlying content is not encrypted, and software that chooses to ignore the flags can. A permissions password discourages casual copying; it does not prevent a determined reader.

Removing protection you are entitled to remove is a routine operation, not a workaround — a document you own, with a password you know, that has become inconvenient to keep re-entering.Unlocking a PDF you have the password for covers that case, which is the only one these tools support: nothing here recovers or bypasses a password you do not have.

Two further limits are worth naming, because they are where expectations usually break. The first is scope. Encryption protects the file, not the information in it. Once a recipient has opened the document they can copy from it, photograph the screen, or forward the file and the password together, and nothing in the format prevents any of that. A password controls who gets in; it has no reach past the moment someone does.

The second is ordering. Every other operation needs the document decrypted first, so a file that is protected early has to be unlocked again before it can be merged, cropped or numbered. Protection belongs at the end, once the document is finished — otherwise the work gets done twice, and each round trip is another chance to save the wrong version.

There is also no recovery, and that is a feature rather than a shortcoming. A correctly encrypted PDF holds no record of its own password anywhere in the file, so no tool and no support process can retrieve one. Write the password down before closing the tab, and send it to the recipient through a different channel than the document — a password sitting in the same email thread as its attachment is protecting nothing.

When a PDF is too large, and what actually shrinks it

File size complaints almost always trace back to one of a small number of causes, and identifying which applies decides whether the document can be reduced meaningfully or barely at all.

The dominant cause is images. A text-only document of many pages is typically small, because text is stored compactly as characters and font references. A scanned document of the same length can be many times larger, because every page is a photograph. When a file seems disproportionately heavy for the number of pages it has, scanned or embedded imagery is nearly always the reason —why a PDF gets so large walks through the diagnosis.

Two other contributors are easy to overlook. Fonts embedded in full rather than subset can add substantial weight, particularly when several typefaces are involved. And material that has accumulated across earlier edits — previous revisions, unused objects, annotation history — can remain in the file long after it stops being visible.

What actually reduces size follows from the cause. Recompressing images is the lever that moves scanned documents, and it trades a degree of visual fidelity for the reduction. Removing pages you do not need genuinely helps, because their content leaves with them. Cropping does not, for the reason given further up this page: the material outside the crop box is still in the file. And when a document simply cannot be made small enough, sending it differently is often the better answer —a PDF too large to email covers the practical options, including splitting a long document into parts that each fit.

Two things reliably do not help, and both get tried first. Zipping a PDF is the common one: the images inside it are already compressed, and a second pass over compressed data finds almost nothing left to remove, so a four-megabyte document becomes a four-megabyte archive with an extra step for whoever receives it. Saving a copy under a new name is the other — it rewrites the same objects and changes nothing that counts.

Where the size came from also decides where it can be fixed. A page captured at a resolution far higher than the document will ever need is heavy before any tool sees it, and recompressing afterwards cannot recover what a sensible capture setting would simply never have created. If the paper is still to hand and the batch is large, rescanning is often faster and produces a better document than compressing a bad scan.

Failing that, the order that works is to remove what is not needed, then recompress the images, and only then consider dividing the document or sending it as a link rather than an attachment. Doing it the other way round — compressing hard first — degrades pages you were about to delete anyway, and leaves less room to manoeuvre if the result is still too big.

Marking up a document: text, watermarks, page numbers

Three tools here add something to a page rather than changing what is already on it. They work in similar ways and are often used in sequence, but each answers a different need.

Adding text places new words at a position you choose. It is what you want for filling in a form that was never made fillable, correcting a detail, or annotating a page for someone else to read. The text sits on top of the existing content rather than flowing with it, which is why it does not reflow if the surrounding layout differs — adding text to a PDF covers placement and the cases where a form field would serve better.

A watermark is text or a mark repeated across pages to say something about the document's status rather than its content: draft, confidential, superseded. Because it applies to every page rather than a chosen spot, it communicates at a glance even when a single page is printed and separated from the rest. Adding a watermark covers placement and opacity, both of which decide whether it reads as a status marker or an obstruction.

Page numbers serve documents that will be referred to rather than merely read — anything someone will cite, discuss in a meeting, or file. They are worth adding last, because numbering is derived from the order the pages are currently in. Combining files or rearranging them afterwards leaves the printed numbers disagreeing with the actual sequence, which is more confusing than having no numbers at all. Get the order settled first — reordering pages before numberingexplains why this sequence matters — then useadding page numbers as the final step.

Before you send it: a short pre-flight check

Most problems with a finished document are discovered by whoever receives it. A brief check before sending catches nearly all of them, and none of the steps takes more than a moment.

1. Open the exported file, not the original. Confirm you are looking at what you actually produced rather than the document you started from.
2. Check the page count and order. Particularly after merging or splitting, where an off-by-one is easy and invisible until someone reads it.
3. Try selecting a line of text. If nothing selects, the document is a scan, and the recipient will not be able to search it either.
4. Confirm page numbers match the sequence if you added them before making further changes.
5. Look at what the file carries beyond its pages — author names, software details and revision traces travel with a document unless removed. What hidden data a PDF carries covers what is stored and where.
6. Check the size against how you are sending it before the message bounces rather than after.

For documents going somewhere formal — a court bundle, a tender, a submission with stated requirements — the order in which these steps are performed matters as much as performing them.Preparing a PDF bundle for submissionsets out a sequence that avoids having to redo earlier work.

Two of these checks get skipped far more often than the rest, and they are the two that cause the most rework. The first is opening the exported file rather than trusting the preview: a preview shows what the tool intends to write, and the file shows what it actually wrote. The second is looking at it somewhere other than where you made it. A document can render correctly in the browser that produced it and badly in the recipient's reader, usually because a font the file expects is not embedded and a substitute is standing in — which changes spacing and can push a line of text out of position.