Combining documents without rebuilding them
Merging PDFs is one of the few document operations that is genuinely lossless. Pages are copied from one file into another whole — the text stays text, the images stay at their original resolution, the fonts come along embedded, and nothing is re-rendered or re-compressed on the way.
That is worth knowing because the alternative people reach for is not lossless at all. Printing several documents to a single PDF, or exporting them through an image step, flattens text into pictures of text and inflates the file. The result looks similar and is a much poorer document.
Order is the whole job
Almost every merge problem is an ordering problem. Files arrive named by whoever made them, and the order they appear in a file picker is the order the operating system chose, which is rarely the order you meant.
Two habits fix it permanently. Name files by the order you want them in, with a leading number and a leading zero — 01-cover, 02-summary, 10-appendix — because without the zero, ten sorts before two. And check the sequence in the list before merging rather than in the output afterwards, which is faster than merging twice.
The PDF zusammenfügen works on files you select together and takes at least two, which is the only real constraint. Everything happens in your browser, so nothing is uploaded anywhere — worth caring about, since the documents people merge are contracts, invoices and scans of things they would rather not post to a stranger’s server.
What travels with the pages
More than the visible content, and this occasionally matters:
- Embedded fonts come with the pages, which is why merged documents can end up carrying several near-identical font subsets and growing more than the arithmetic suggests.
- Annotations and form fields generally survive. Form fields sharing a name across two merged documents will interfere with each other, because a PDF form treats same-named fields as the same field.
- Page size and orientation stay as they were. Merging A4 and Letter produces a document containing both, which is legal, and which a printer will handle by scaling or clipping.
What does not survive is document-level structure: bookmarks, the outline tree, and the metadata of every file except the one the merge is built on. A long merged report usually wants its bookmarks rebuilt afterwards if anyone is expected to navigate it.
Size is not the sum of the parts
A merged file is often smaller than its inputs added together, because shared resources are stored once. It is sometimes larger, because each source contributed its own font subsets and colour profiles.
Either way, do not be surprised by the number. If the result needs to fit an upload limit, that is a separate operation performed after merging, not something the merge itself controls.
When not to merge
Two cases come up repeatedly and both are worth resisting.
Email attachments. Merging twelve invoices into one file makes them easier to send and much harder to use, because the recipient wanted to file them separately and now has to split them again. Send them as they are, or as a zip.
Documents with different lifespans. A merged pack has to be rebuilt entirely when one component changes. If the terms document is revised quarterly and the covering letter is not, keeping them separate saves the quarterly rebuild.
Merging is right when the result is genuinely one document — a report assembled from sections, a submission with appendices, a scanned set of pages that were always one thing.
Scanned pages, which is the common case
A great deal of merging is putting a document scanned in pieces back together, and the pitfalls are practical rather than technical.
Pages scanned in two batches often differ in orientation, because the second batch went into the feeder the other way. Check the output for upside-down pages before sending it. Double-sided originals scanned as two passes need interleaving rather than appending, which no merge does for you — the front pages and the back pages have to be ordered into place first.
And a scan is images, so a merged scan is a large file with no searchable text at all. That is inherent to the scan rather than to the merge, and it is why a fifteen-page scanned pack can outweigh a two-hundred-page typeset report.
Check the seams
Open the result and look at the joins — the last page of each source and the first page of the next. That is where the ordering mistakes, the duplicated cover pages and the missing final page all show up, and it takes about as long as scrolling.