Merge PDF

What a merge keeps, and what it quietly drops

No merging tool can avoid this entirely, because two documents cannot both keep their own document-level furniture inside one file. It is worth knowing which parts you are losing.

Page by page, into an empty document

Nothing appeared in the list

Only files the browser reports as PDFs are accepted. Anything else is skipped, and a short message says how many files were skipped. Two common causes:

• It is not really a PDF. A .doc or a renamed file will not be read. Export a real PDF from the original application.
• It arrived with no file type attached. Some sync folders and archive extractors hand over files with the type field empty. Save the file to your desktop first, then add it from there.

Also note that the merge button only appears once there are two or more files in the list, and that adding the same file twice puts it in twice. That is occasionally useful — a cover sheet repeated — and occasionally a surprise.

Getting the sequence right

Files merge in the order they appear in the list, and the list is the order you added them. Each row has ↑ and ↓ buttons to move it, and a Remove button.

In practice the reliable method is to add files one at a time, in the order you want them, rather than selecting twelve at once and hoping the file picker agrees with you about alphabetical order. If a document lands in the wrong place, move it with the arrows.

If what you actually need is a different page order within one document, that is a job for split PDF: list the pages in the order you want them and it rebuilds the file that way.

Memory is the only ceiling, and it is a real one

There is no file count limit and no page limit, because there is no server imposing one. What there is instead is your browser tab. Every file is read fully into memory, all of their pages are copied into the new document, and then the whole output is assembled as one block of bytes before the download starts. Peak memory is roughly everything at once.

On a desktop that is rarely a problem. On a phone, merging a stack of large scans can stall or fail outright. When it does:

1. Merge two or three files, download the result.
2. Merge that result with the next two or three.
3. Repeat. The peak stays low even though the final document is large.

A second thing to expect: the total page count is recalculated by re-reading every file each time the list changes. With a dozen big documents the counter takes a moment to settle. It is counting, not hanging.

Password-protected files will not open here at all, and a file the parser considers malformed will fail even if a desktop reader opens it happily. Re-export a clean copy from a reader first.

Assembling a rental or mortgage application

The classic case: the agent wants one PDF containing ID, three payslips, two bank statements and a reference letter, and half of those exist only as phone photos.

1. Turn the photos into PDFs first with images to PDF, one document per item.
2. Trim the statements down to the pages that were asked for using split PDF. Sending twelve months when three were requested is its own kind of oversharing.
3. Add everything here in the order the agent listed it, then merge.
4. If the result is too large to email, run it through compress PDF — though if it is mostly photographs, expect little, and shrink the images before step one instead with compress image.

The most sensitive pile of paper you own

Merging is unusual among file tasks: the files you combine are almost always related, and related documents tell a much fuller story than any one of them. Passport, payslips and bank statements uploaded separately are three leaks. Uploaded as one merge job they are a complete identity package, assembled and labelled for whoever ends up with it.

That is why this runs where it does. The library executes in the tab, your files stay on your disk, and the merged document is written straight to your downloads. There is no upload to interrupt, no queue, and nothing left behind to delete. It also keeps working with the network off, which shows the work does not need a server; the Network tab shows what, if anything, is sent.

Everything in the tool list follows the same rule, including the image tools.