How to split a PDF

The three ways, and what each is for

By range

You name the cut points: pages 1–4, 5–12, 13–20. This is the right choice when the document has a structure you already know — chapters, sections, one invoice after another in a batch export. Ranges can also be fixed: every 10 pages, without you listing them, which suits a document with a regular shape such as a scanned book or a monthly report.

There is one option worth understanding here. Normally each range becomes its own file. If you ask for the ranges to be merged instead, you get a single PDF containing only those pages — which is really an extraction, not a split, and is how you drop the pages you do not want while keeping the rest together.

By pages

Every page becomes its own PDF, or only the pages you list do. A 40-page document split this way produces 40 files. That is genuinely what some jobs need — separating a stack of scanned forms, or pulling three specific pages out of a long report — and it is almost never what someone wants by accident, which is why a tool should tell you how many files you are about to get before it makes them.

By size

You give a limit — 5 MB, say — and the cuts land wherever they have to in order to keep every piece under it. This is the one people reach for when a mail server has rejected an attachment, and it is the only one of the three where the page numbers are not the point.

Choosing between splitting, extracting and deleting

Three tools overlap here, and the difference is what you get back.

• Split PDF gives you several files from one.
• Extract pages gives you one file containing a selection.
• Delete pages gives you one file with a selection missing.

Extract and delete are complements of each other: extracting pages 2–5 and deleting pages 2–5 produce the two halves of the same document. If you want both halves, splitting by range is one operation instead of two.

Going the other way, merge reverses a split. Splitting and re-merging is lossless in the sense that matters — the page content is copied, not re-rendered — though see the caveat below about what does not survive.

Why splitting to a size is not exact

A PDF is not a stack of independent pages, and this is the single most useful thing to know about splitting one. Fonts, embedded colour profiles and repeated images are shared objects: the file stores each one once and every page that uses it points at the same copy. Splitting has to give each output file its own copy of everything its pages reference.

Two consequences follow, and both surprise people:

• The parts can add up to more than the whole. Split a 10 MB report with one embedded font family into five files and you may get five files totalling 12 MB, because that font is now embedded five times.
• The size of a piece cannot be predicted from its page count. Ten pages of text and ten pages of photographs are not comparable amounts of data.

This is why an honest size split has to measure rather than estimate: add a page, write the file, see how big it actually became, and close the file before the next page would push it over. iBuildPDF does exactly that, which is why size mode takes longer than the other two on a large document.

It also means a limit cannot always be met. A single page containing a high-resolution photograph may be larger than your limit all by itself, and no amount of cutting will change that — a page cannot be split in half. When that happens the file is still produced and you are told, rather than being handed something quietly over the limit. Compress PDF first is usually the answer.

What to check afterwards

Splitting copies pages faithfully, but a PDF carries things that belong to the document rather than to any page, and those do not survive being divided:

• Bookmarks and the outline point at pages that are now in different files.
• Internal links — a cross-reference to page 40 — have nothing to point at once page 40 is in another document.
• Form fields that were part of one form may be separated from the rest of it.
• Digital signatures are invalidated. A signature certifies a specific byte sequence; any split produces different bytes, so the signature no longer verifies. This is the signature working correctly, not a fault in the tool.

Page numbers printed on the pages themselves are just ink and will still say what they said. If the new files need their own numbering, add page numbers after splitting, not before.

A quick check with page count and file info on each output confirms you got the pages you expected and how large each piece ended up.