Taking pages out without degrading them
Splitting a PDF copies whole pages into a new document. Nothing is re-rendered, so the text stays selectable, images keep their original resolution, and the extracted pages are byte-for-byte as good as they were.
This is worth contrasting with how people usually do it. Printing a range to a new PDF works, and on many systems it also flattens the text, drops the annotations and re-compresses the images. Screenshotting pages is worse again. Splitting properly costs nothing in quality; the alternatives quietly cost a great deal.
Why you would split rather than send the whole thing
Usually because the recipient should only see part of it. A single clause from a contract, one invoice from a monthly batch, the chapter somebody asked about rather than the four hundred pages around it.
Sometimes because of a limit — an upload cap, an email attachment ceiling, a portal that refuses anything over a certain size. And sometimes because a document that arrived as one file was never really one document: a batch scan of twenty separate forms is twenty documents that happen to share a file.
Writing a page range
The [msm_tool slug="pdf-split"] takes a range in the ordinary form — 1-3, 7, 9- for the first three pages, page seven, and everything from nine to the end. Leaving it blank takes every page.
Two things catch people out, and neither is the tool's doing.
Printed page numbers are not PDF page numbers. A report with a cover, a contents page and roman-numbered front matter has a page marked 1 that is the ninth page of the file. Ranges count pages in the file, from one, ignoring whatever is printed on them. Scroll to the page you want and read the position from the reader's page indicator rather than from the page itself.
Ranges are inclusive at both ends. 1-3 is three pages, not two. Obvious written down, and the source of a great many extractions that are one page short.
One file per page
The other mode splits a document into a separate PDF for each page, delivered as a zip. It is the right answer for a batch — a scanner run that produced fifty separate certificates, a payroll export containing one payslip per page, a set of forms that need routing to different people.
It is the wrong answer for a report, which then becomes two hundred files nobody can read in order. If you find yourself about to page-split something with continuous prose, you probably wanted a range.
Filenames from a per-page split are positional rather than meaningful. If the pages are going to different people, renaming them is the real work, and it is worth thinking about before generating fifty files named after their index.
What splitting does not do
It does not remove information from the pages it keeps. This is the important one. Extracting pages four to six of a confidential document gives you three pages that still contain everything those three pages contained, including anything hidden under a black box, in the metadata, or in an annotation. Splitting is not redaction and it is not sanitisation.
It does not renumber anything. Page numbers printed on the pages come along as they are, so an extract of pages 40 to 45 opens at a page marked 40. That is usually correct behaviour — the extract is honest about where it came from — but it surprises people who expected a document starting at one.
It does not fix a broken source. A password-protected or damaged file has to be dealt with before it can be split rather than after.
Bookmarks, links and forms
An extract loses the document outline, because a bookmark tree describing four hundred pages means nothing across six. Internal links pointing at pages that are no longer in the file become dead, which matters most in documents with cross-references.
Form fields survive on the pages that carry them, but a form split across pages becomes several partial forms, each independently fillable and none of them the original. If a form has to be completed, keep it whole.
Check what you actually extracted
Open the result and read the first and last pages before sending. Off-by-one at either end is the standard mistake, and the version where you sent one page more than you meant to is the one worth catching — particularly when the reason for splitting was that the recipient should not see the rest.