Merge PDF files

The locked file that comes back empty

There is a failure in this category that almost nobody talks about, and it is worth understanding before you trust any tool with a contract. Most PDF libraries cannot decrypt a protected document. What they offer instead is a flag that skips the encryption check — not the encryption itself. Turn it on and the file loads without complaint, the pages are all counted, and the merge appears to succeed. The content streams, meanwhile, were copied across still scrambled, into a container that now claims not to be encrypted at all.

The document you download from that process opens fine. It has the right number of pages. Every one of them is blank. We measured this before writing a line of this page, on a test file containing a single known sentence, and the extracted text came back as an empty string. Nothing in the chain objects, which is the dangerous part: the tool does not know it failed, so it cannot tell you.

This page will not do that. A protected file is named as locked and given a password box on its own row, and nothing is taken from it until the password actually opens it. If you do not have the password, the file stays out and the merge button stays disabled rather than quietly producing a document with hollow pages in the middle. The material people bring to a merge tool is contracts, invoices, transcripts and medical records. A tool that hands back an empty agreement is worse than one that admits it cannot help.

If the same file keeps arriving locked and you would rather deal with the lock once instead of on every merge, removing the password from the PDF is its own page. It is also the place to find out which of the two PDF passwords you are actually facing, since one of them can be lifted without knowing any password at all and the other cannot be lifted without it.

PDF, in brief

PDF

Full name	Portable Document Format
Extension	.pdf
Format type	Fixed-layout document
MIME type	application/pdf

When you merge PDF files, order is the entire job

Combining two files is trivial. Combining eleven in the right sequence is the actual task, and it is where the time goes. A signed agreement followed by its appendices, a set of scanned invoices for one quarter, a report whose chapters were written by different people and exported separately — in every case the pages themselves are fine and the sequence is the thing you are being paid to get right.

The list is therefore the instrument, not a summary of one. Every row carries a grip that responds to a pointer, and on a touchscreen to a finger held still for a moment before it slides. Beside the grip sit two arrows and a discard control, each of which responds to one tap and nothing else. This redundancy is not politeness. Sliding a row precisely is genuinely difficult on a phone, and a tool whose only reordering gesture requires that precision simply does not function on the hardware most people are holding. Rearranging touches nothing on disk, so exploring a different sequence is free.

A file row is still too coarse when page two belongs after a sheet from another file. The numbered picture strip below the rows is the actual output order, not a decorative gallery. Every page has earlier and later buttons with full touch targets, and a move there changes the byte order in the PDF that gets written. The pictures are drawn only when they approach the screen, so a long report is not forced into hundreds of canvases merely because it was opened. The page box still controls which sheets are included; typing 2, 1 there is a selection, not a hidden ordering command. The visible picture strip is the one place that controls sequence.

Small pictures make it hard to tell two similar cover sheets apart. Tap any page in the strip and it fills the window, rendered fresh from the document so the lettering stays sharp instead of being a stretched copy of the thumbnail. Previous and Next, or the arrow keys, walk through the combined file in its final sequence. After the merge, the receipt spells out that same order line by line — which output pages came from which file — so a sheet moved past another document is visible on a phone too, where the inline viewer is not drawn.

Before you run anything, each row tells you how many pages the file has, how large it is, and what size its pages are. That last one catches a real problem: a batch that mixes A4 and Letter merges perfectly well, because every page keeps the dimensions it already had, but it prints on mixed paper and you would rather learn that now than at the printer. The estimated size of the finished document sits above the button too, not after the work, so nothing about the outcome is a surprise.

Taking only some of the pages

Every row has a page box that starts out saying All pages, and most of the time that is the correct answer and you can ignore it entirely. When it is not, the box accepts the forms people actually type rather than one rigid syntax: 1-3, 7 for a list, 5- for everything from page five onward, -4 for everything up to page four, and last when you want the final page without counting to find out which one it is. It also quietly accepts the en dash that word processors substitute while you type, because pasting a range out of a document should not produce an error.

If you name a page the file does not have, it says so — which page, and how many the file actually contains — instead of handing back fewer pages than you asked for and letting you discover the gap later. That combination is why the box exists at all: taking pages one to three from one file and four to nine from another is a split and a merge in a single pass, and doing it here saves a round trip through a second tool and a download you did not need.

The box is the wrong shape for one common job, though, and it is worth saying so. If what you want is a document minus two of its forty sheets, naming the thirty-eight survivors here is tedious and easy to get subtly wrong. Deleting pages from a PDF puts the sheets on screen as thumbnails and lets you point at the two, which is the same result reached by an instruction you can check at a glance.

Nothing is re-drawn on the way through

Pages are carried across exactly as they already are. Same dimensions, same fonts, same images, no re-compression and no re-rendering. A page that was a crisp vector diagram stays a crisp vector diagram, and text that could be selected and searched before can still be selected and searched afterward. That last property is easy to lose: any tool that rebuilds pages by rasterizing them turns a searchable document into a stack of pictures, and the damage is not visible until someone tries to copy a paragraph out of it. To audit what came through intact, PDF to text extracts the text layer for inspection, while how to copy text from a PDF sets out why some files carry that layer and others never possessed one.

There is one genuine casualty, and it is worth being precise about because the imprecise version is scarier than the truth. A fillable form stops being fillable. A PDF form does not live on the pages; it hangs off the document as a separate index of fields, and combining documents means constructing a new document, which that index does not travel to. We measured what actually happens: the merged file reported zero form fields, while the name that had been typed into it was still there on the page. In other words the answers survive as ink and the form does not survive as a form — nothing you filled in disappears, but nobody can edit it afterward, and software that reads submissions by field name will find none. Any row carrying fillable fields is flagged with a count before you run anything, so this is never a discovery you make after sending the file. If the document still needs to be filled in later, merge first and fill second. Bookmarks and the document outline are lost outright, for the same structural reason — that was measured too, and unlike the form values they leave nothing behind.

This also means there is no page size setting here, and its absence is deliberate rather than missing. Rescaling pages to a common size would mean redrawing them, and a contract whose dimensions changed is not quite the same document anymore. Merging and resizing are different jobs, and mixing them into one button is how tools end up doing both badly.

What this page does not do

It only combines. Splitting one document into several is the opposite direction and lives on its own page rather than as a toggle here, so that a link to a merge tool always lands on a merge tool — split PDF pages is where that job lives instead. It will not compress the result, so a merge of large scanned files produces a large file — that is arithmetic, not a flaw. It will not add page numbers, bookmarks or a table of contents. And it will not perform character recognition on scanned pages, so a merged stack of scans is still image-only, exactly as the originals were. Nor does it straighten anything: a page that arrived on its side is combined on its side, since each page keeps the orientation entry it already carried. That is usually what you want, but if one of the stacks came off a scanner sideways it is worth rotating those pages first, because the turn is written into the file and therefore survives being merged.

There is, however, one thing it categorically will not do, and that is turn you away. No ceiling applies, on either file size or page count. Hand it something enormous and you get a straightforward note about the memory the operation is likely to consume, followed by the operation. The hardware belongs to you and so does the decision. Documents that turn out not to be PDFs, and PDFs too corrupted to parse, are listed by name and given distinct explanations, since they call for entirely different remedies — and the remainder of the batch proceeds untouched. Should your starting material be pictures rather than paperwork, JPG to PDF and PNG to PDF build a document from photos or screenshots directly, which is one step instead of two.

Nor does it sign anything, and that is worth mentioning because merging is so often the step before it. A contract that arrives as a cover letter, an agreement and two exhibits becomes one document here, which is the version worth signing: one file, one signature, one thing to send back. Signing a PDF is the next page along, and it edits the merged document in place rather than rebuilding it, so everything this page just preserved stays preserved.

A promise you can check instead of believe

Plenty of sites in this category say the words in your browser. It is a claim worth treating with suspicion, this page included, since the sentence costs nothing to write and the truth of it cannot be seen from outside. So verify it. Bring up the network panel in your browser's inspector and run a merge while watching: this page requests its own code and the self-hosted PDF preview worker after the first file is handed in, and after those local assets arrive the panel stays quiet. No request carries the document bytes.

The reason it can work this way is worth stating plainly, because it is mechanical rather than a policy we could quietly change. The page pictures are rendered locally for inspection, but combining the file itself is not image work: a PDF stores its pages as a tree of numbered objects, so the writing library parses that tree straight out of the bytes your browser read, lifts the page objects in the visible order, and writes them into a freshly built document. The preview never replaces the source page, so text and vectors stay text and vectors. A server would add latency, a queue, and a copy of your contract on someone else's disk, in exchange for nothing.

Compare that with a service promising to erase your uploads within the hour. Ours is the stronger guarantee for an unglamorous reason: material that was never transmitted cannot subsequently be retained, breached, subpoenaed, or handed to some processor you never agreed to. A useful side effect is that once this page has loaded it keeps working with the network switched off entirely. For a fuller account of how to audit this claim anywhere, not merely here, see how to check that a site is not uploading your files.