Splitting a PDF: single pages, page ranges and fixed-size chunks
There are only three useful ways to split a PDF: one file per page, a list of ranges that you choose by hand, or equal batches of N pages. The right choice follows from who receives the parts and whether the document has internal structure worth preserving. Everything else in this guide is about the details that decide whether the parts arrive in a usable order.
The three modes, side by side
| Mode | Produces | Best for | Weak at |
|---|---|---|---|
| One file per page | N files, one page each | Forms, coupons, single-page letters to different people | Anything with a running header or shared figure |
| Page ranges | One file per range you type | Chapters, signed sections, extracts for a specific reader | Long lists of ranges on a 400-page file |
| Every N pages | ceil(N_pages / size) files | Email attachments, upload limits, print batches | Documents whose sections do not align to N |
One-file-per-page is the mode people reach for first and regret most often. A page lifted out of a report loses the context that the header, the numbering and the preceding page provided. If the recipient needs one page, that is fine. If they need one section, ranges are almost always the better answer.
The arithmetic of fixed chunks
A document of P pages split at N pages per file produces ceil(P / N) files. The last one holds P mod N pages, or N pages if the division is exact. A 32-page report at N = 10 gives four files: 10, 10, 10 and 2. The 2-page tail is not padded, and that is deliberate - padding would add blank pages that later fail a page-count check.
Two practical consequences follow. First, choose N so that the tail is small: at N = 10, a 32-page file has a 2-page remainder, which is fine; at N = 7, the tail is 4, and at N = 30, the tail is 2 but the first file is 30 pages, which may exceed the email limit you were trying to satisfy. Second, never split a document with numbered sections at a fixed interval without checking where the cuts land. A chapter that starts on page 21 and ends on page 34 will be cut in half by N = 10.
Range syntax and the four ways it fails
A range list is comma-separated, and each part is either a single page or a low-to-high span. 1-3,5,7-9 means three files: pages 1 to 3, page 5, and pages 7 to 9. The failures are predictable:
- Reversed spans.
9-7is not "pages 7 to 9". It is an error, and treating it as a valid range is how people end up with pages in the wrong order. - Out-of-range numbers. Asking for page 40 in a 12-page file should stop the run, not silently produce anything. A tool that clamps to the last page gives you a file that looks right and is not.
- Page labels instead of numbers. Many documents show "iii" or "A-4" in the footer. Those are page labels, not page indices; the split field wants the physical position, which you can read from the page counter in your reader.
- Pasted typography. Word and Excel paste en dashes and non-breaking spaces. A range field that only accepts the ASCII hyphen will reject
1–3without saying why, which reads as a broken tool.
Overlapping ranges, on the other hand, are legitimate: 1-3,2-4 produces two files that share pages 2 and 3. That is useful when two people need overlapping extracts and it is not a mistake worth blocking.
What splitting does to the file structure
Splitting is the reverse of merging, but it is not symmetric in cost. Each output document needs its own copies of the fonts, images and colour profiles that its pages use, because a PDF cannot reference resources that live in a different file. A 40-page catalogue split into 40 single-page files can end up several times larger in total than the original, while a split into four 10-page parts barely grows at all. If total size matters, split less finely.
Internal links behave differently too. A link from page 2 to page 30 works if both pages land in the same part and breaks if they do not. Most splitters do not rewire cross-part links, and it is not reasonable to expect them to: there is no target to point at. When you split a document that relies on internal navigation, either keep the linked pages together or accept that the links will be inert.
Bookmarks follow the same logic. An outline entry that points inside the extracted range can be preserved; one that points outside has nowhere to go. Many tools drop the whole outline rather than keep half of it.
Naming that survives a folder sort
File names are sorted as text by every file manager, every email client and every script. That makes two rules non-negotiable:
- Zero-pad any counter.
part-01topart-12sorts correctly;part-1topart-12sorts as 1, 10, 11, 12, 2. The same applies to page numbers in per-page mode. - Put the sortable part first. A name like
report-part-03-pages-21-30.pdftells you the document, the position in the sequence and the pages it holds, in that order. A name likepages 21-30 (report) final v2.pdfis readable to a human and useless to a sort.
Dates belong in ISO form, year first: 2026-11-19. Local formats such as 11/19/26 sort wrongly and are ambiguous to anyone outside the country that uses them. If a version matters, put it after the sequence number, not in the middle of the name.
Checking a split
A split is easy to verify because the arithmetic is exact. The page counts of the parts should add up to the page count of the source, minus any pages you deliberately did not include. If the total is short, a range was mistyped. If it is long, ranges overlapped. Both take one addition to spot:
| Source | Ranges | Expected parts | Expected pages |
|---|---|---|---|
| 12 pages | 1-3,5,7-9 | 3 | 3 + 1 + 3 = 7 |
| 12 pages | 1-6,7-12 | 2 | 6 + 6 = 12 |
| 32 pages, N=10 | fixed chunks | 4 | 10 + 10 + 10 + 2 = 32 |
| 12 pages | 1-3,2-4 | 2 | 3 + 3 = 6, with pages 2-3 duplicated |
Open one part from the start of the document and one from the end. Check the page in the top corner and the footer. If a running header says "Page 7 of 12" on the first page of a part, the split is correct - the printed numbers are content, not structure, and they should not change.
Splitting as the first half of another job
Half the time, splitting is not the goal but a step. Extracting one section to send to a reviewer, pulling a single signed page out of a contract, or reducing a file to a size an upload form accepts are all split-then-send jobs. The other common pattern is split, edit, merge: take a document apart, replace two pages, then reassemble it in the merge tool with the original order. That workflow is safer than editing the whole file in place, because each part can be checked before it rejoins the document.
When the parts need numbers of their own - a set of extracts that will be bound together later - number them after splitting rather than before. Renumbering a part that already carries the source document's footer produces two numbering systems on one page.
Frequently asked
Can I split a PDF by bookmark or by chapter automatically?
Not in this tool, and rarely well anywhere. Splitting at bookmarks requires the outline to point at real page objects in a consistent order, which is common in professionally produced files and rare in exports from office software. Check the outline in your reader; if the entries resolve to the pages you expect, a desktop editor can split at those points.
Why are the parts larger than the original in total?
Shared resources are duplicated per file. Fonts, images and colour profiles have to be present in every document that uses them. Fine splits multiply that overhead; coarse splits barely feel it.
Does splitting remove the password from an encrypted PDF?
No. Encryption travels with the pages. A file that requires a password to open still requires it after splitting, and extracting pages from a restricted file does not grant rights you did not have.
Related reading: page order and bookmarks when merging and archive naming for contracts and invoices.
Written by the PDFNest team. Last updated November 19, 2026.