Splitting a PDF is usually a practical necessity rather than a preference: a bank wants only the statement pages, a portal caps uploads at ten pages, or a hundred-page scan contains one contract you actually need. What gets carried across, and what quietly does not, is worth understanding first.
Extraction should be lossless
A correct split copies the page objects — text, fonts, vector paths, and images — into a new document without re-rendering. The extracted pages are byte-for-byte equivalent in content: text stays selectable, images stay sharp, and the file stays searchable.
Tools that re-render pages as images during extraction destroy all of that. The tell is the same as with merging: if you can no longer select text in the output, it was rasterized. This also breaks accessibility, since screen readers have nothing left to read.
What does not survive extraction
Several document-level features are attached to the whole file rather than to individual pages, and they can be lost or broken by splitting.
- Bookmarks and the outline tree, which point at pages that may no longer exist in the extract.
- Internal cross-reference links, which break if the target page was not included.
- Form fields, which often depend on document-level structures and may not carry across intact.
- Digital signatures, which are invalidated by definition — a signature attests to the whole document, so a subset cannot remain signed.
That last point matters practically: extracting pages from a digitally signed contract produces a document that no longer proves anything about its own integrity.
Page numbers are printed, not computed
A page number visible on the page is usually drawn into the content, not generated dynamically. Extract pages 40 to 45 and the output will still show 40 to 45, even though it is now a six-page document. This confuses recipients, so add a note or re-number if the extract will be circulated.
The related trap: front matter often means the printed number does not match the position in the file. Page 1 of a report might be the fifth page of the PDF.
File size does not fall proportionally
Extracting ten pages from a hundred-page document rarely produces a file a tenth of the size. Embedded fonts are document-level and get carried over largely intact, and some tools retain unreferenced objects. If the extract still exceeds an upload limit, compressing it afterwards is the usual fix.
Encrypted files must be unlocked first
A password-protected PDF cannot be split until it is decrypted, because the tool cannot read the page content. Open it with the password in a viewer, save an unprotected copy, then split that — and delete the unprotected copy afterwards if the content is sensitive.