Split PDF — extract pages as separate files

What happens when a PDF is split

A split walks the input's page tree, copies the chosen pages into a new document along with the resources those pages reference (fonts, images, color profiles, embedded objects), and writes a fresh cross-reference table. Like merging, the operation is byte-level — pages are not re-rendered, so the output looks identical to the matching pages in the source. Form fields, annotations, and bookmarks that fall inside the chosen page range carry over; ones that pointed elsewhere are dropped (because they have nowhere to land).

When you split into multiple output files at once we package them into a zip with predictable filenames — by default 'original-name-1.pdf' through 'original-name-N.pdf', or you can choose a pattern that includes the page range or a section title pulled from the document's bookmarks.

Which split mode to use

Pick the mode that matches your end state, not the most flexible one. Each mode has a job it does well and a job it does badly:

• Each page as a separate file. Use when downstream systems expect single-page documents — for example invoice processing, OCR pipelines that batch by page, or signature workflows that route page-by-page.
• Custom ranges (1-3, 5, 7-9). Use when extracting a known section from a known location — a chapter, an appendix, a specific exhibit. Best mode when you know exactly which pages you want.
• Every Nth page. Use for booklet imposition (every 4th page = every cover quarter) or to sample a large document.
• By file size. Use when the destination has a per-file cap — email gateways at 25 MB, court e-filing portals at 35 MB, regulatory submissions at 100 MB. We iterate to land just under your target.
• By bookmark / heading. Use to break a long document into one file per chapter or section automatically. Requires the source PDF to have a bookmark structure — most exported reports do.

Common mistakes when splitting

These are the patterns that cause people to redo a split or end up with files they can't actually use:

• Splitting before OCR. If the original is scanned, OCR the whole document first. Splitting destroys the cross-page text continuity that improves OCR accuracy on hyphenated words and headers/footers.
• Using ranges that overlap accidentally. 1-5, 4-10, 9-15 will produce three files with duplicated pages. Our parser flags overlaps before running — read the warning.
• Forgetting that pages start at 1, not 0. Every PDF tool worth using uses 1-based page numbers because that's what humans see. If your script-based tool returns pages 0-4 for 'the first five', you'll be off-by-one until you adjust.
• Splitting password-protected PDFs without the password. We can read most encrypted PDFs without a password if the only restriction is 'no extracting' (which we honor only when print restriction is also set). When the file is genuinely password-locked, you'll get an error — unlock it first.
• Trusting bookmark splits on documents with no bookmarks. Many scanned and exported PDFs have no real outline. We fall back to a regex on heading text, which is best-effort. For surgical splits, define ranges manually.

Tips for clean output

Small choices in the split dialog change the quality of the result more than people expect:

• Use a smart filename pattern. {name}-pages-{from}-{to}.pdf is more useful than {name}-1.pdf when the recipient has to find a specific section months later.
• Ask for bookmarks-preserved. On range splits, the option 'preserve outline pointing inside the range' keeps cross-references that land in the output and drops the ones that don't. Default behavior is correct for most cases.
• If outputs will be re-merged, keep page sizes. Don't normalize sizes during a split if the next step is re-merging — you'll do the work twice. Normalize at merge time instead.
• Set a target output size on email-bound splits. If you're splitting because a 60 MB PDF won't fit in email, splitting in half by page count rarely lands at 25 MB each. Use 'split by size, target 24 MB' so you actually solve the problem.
• Verify by opening the first and last output. 30 seconds to confirm the boundaries are where you expected them. Splits are cheap to redo.

Limits, formats, and compatibility

Free web tool: input up to 100 MB, output up to 200 files per split. The API and Pro plan have no upper bound; we've split single 12,000-page PDFs cleanly via the batch endpoint. Outputs are PDF/A-1b compatible (or PDF/A-2 if you ask for it), open cleanly in Acrobat, Preview, and every major reader, and preserve digital signatures within their original page range — though signed pages should generally not be split out of context, since the signature's page-bound assertions become harder to verify.