How to merge, split and reorder PDF files

Merging, splitting and reordering are the most common things people need to do to a PDF, and unlike compression they should cost you nothing in quality. If a tool degrades your document while merging it, the tool is doing it wrong.

This guide covers what each operation actually does to the file, and the cases where something can go quietly wrong.

Merging: what a correct implementation does

Merging two PDFs should copy the page objects from each source document into a new one, along with the resources those pages depend on — fonts, embedded images, colour profiles.

Done this way it is completely lossless. Nothing is re-encoded, nothing is rasterised. Text stays selectable, vector graphics stay sharp at any zoom, and image quality is untouched.

You can tell whether a tool did this correctly by checking two things in the output: try selecting text on a page that had selectable text before, and zoom in a long way on a logo or diagram. If the text still highlights and the graphics are still crisp, the merge preserved the document structure. If the text has become unselectable, the tool rasterised your pages and you have lost more than you intended.

One consequence of copying resources properly: the merged file is usually a little smaller than the sum of its parts, because shared resources get consolidated. If your merged file is dramatically larger than the inputs added together, something is duplicating data unnecessarily.

What merging does not preserve

A few things do not survive a merge cleanly, and it is better to know in advance.

Form fields are the big one. If both documents contain fillable forms with fields of the same name, the merged document has a conflict — depending on the tool, fields may be renamed, merged into one, or dropped. If you are combining filled-in forms, flatten them first so the values become part of the page.

Digital signatures are invalidated by definition. A signature certifies a specific document; producing a new document breaks it. This is correct behaviour, not a bug.

Bookmarks, internal links and table-of-contents entries often do not carry across, or carry across pointing at the wrong pages. Many tools simply drop them.

Attachments embedded in the PDF, and page-level annotations in some cases, may not be copied.

For ordinary documents none of this matters. For signed contracts, filled forms, or a long report with a linked contents page, check the output before you rely on it.

Splitting and extracting pages

Splitting is the same operation in reverse: build a new document containing a subset of the pages. Also lossless when done properly.

There are two variants people want. Extracting a range produces one new file containing the pages you selected — useful for pulling a single chapter or a signature page out of a long document. Bursting produces one file per page, which is what you want when a scanner has put twenty separate documents into one PDF.

Page ranges are usually entered as text, and the conventions are near-universal: a hyphen for a range, commas to separate, so "1-3, 8, 11-14" means pages one through three, page eight, and pages eleven through fourteen. Page numbering starts at one and refers to the physical page position, not to any number printed on the page — a document with roman-numeraled front matter will not match.

One caveat: extracting pages does not remove information that lives at the document level. Metadata, and in some cases data embedded elsewhere in the file, can persist. If you are extracting pages specifically to remove sensitive content, verify the output rather than assuming.

Reordering and rotating

Reordering is just building the output document with pages in a different sequence. Nothing is re-encoded.

Rotation is worth understanding because it works differently from how it looks. Rotating a PDF page does not re-render anything — it sets an attribute on the page telling readers to display it turned by 90, 180 or 270 degrees. The underlying content is untouched.

This means rotation is completely free: no quality loss and no change in file size. It also means rotation is cumulative. A page that was already rotated in the source document, rotated another 90 degrees by you, ends up at the sum of both. A tool that sets rotation absolutely rather than adding to what is there will snap already-rotated pages to the wrong angle — a common and confusing bug.

Rotating a scanned page that came in sideways is the usual reason to do this, and it is one of the few PDF edits with genuinely no downside.

Practical notes

A few things that come up repeatedly:

• Order in, order out. Most merge tools combine files in the order you added them, not alphabetically. Check the sequence before merging rather than after.
• Encrypted files must be unlocked first. A PDF that asks for a password to open cannot be read by most tools until the protection is removed.
• Mixed page sizes stay mixed. Merging an A4 document with a US Letter one produces a file where pages are different sizes. That is correct, and it may not be what you wanted — check before printing.
• Keep the originals until you have opened the result and confirmed it. This costs nothing and catches the occasional case where a tool produces a valid file that is not the file you wanted.

A note on where the work happens

Merging and splitting are structurally simple operations that a browser can do perfectly well on its own. There is no technical reason a merge tool needs to upload your documents to a server.

That matters because the documents people most often need to merge — contracts with their signature pages, scanned identity documents, assembled financial records — are frequently the documents they would least like to hand to a third party.

If you want to check whether a tool is uploading, load the page, disconnect from the internet, and try to merge two files. A tool doing the work locally will not notice the network is gone.