How to split a PDF or extract specific pages

“Splitting a PDF” covers three distinct jobs. Extracting pulls one page or a handful of pages out into a new file — the signature page of a contract, one chapter of a manual. Splitting into ranges cuts a document at chosen boundaries — a 90-page scan into three 30-page invoices. Bursting explodes the document into one file per page. All three are the same operation underneath (copy selected page objects into new files), but tools expose them differently, and picking the method by the job saves time.

Method 1: a browser tool that never uploads

The fastest route for most people is a browser-based splitter. Our PDF split tool opens the file with JavaScript in your own browser tab, lets you pick pages or ranges from thumbnails, and writes the output PDFs locally. The document is never transmitted anywhere.

That last property is not a footnote. Contracts, medical records, bank statements, and HR documents routinely get pushed through “free PDF splitter” sites that upload the full file to a server you know nothing about — retention policy, jurisdiction, breach history all unknown. If the document is confidential, the rule is simple: use a tool where the file stays on your machine. In-browser processing satisfies that; you can verify it by loading the tool, cutting your network connection, and splitting anyway. It still works, because there is no server step.

Browser splitters handle extraction, ranges, and bursting equally well, and they work identically on Windows, macOS, Linux, and tablets. Their practical ceiling is very large files — a 500 MB scanned archive can exhaust a browser tab’s memory, at which point the command line (method 4) takes over.

Method 2: macOS Preview thumbnail drag

Preview, already on every Mac, splits PDFs with no extra software. Open the PDF, choose View > Thumbnails to show the sidebar, then:

• Extract pages: select one or more thumbnails (Cmd-click for several) and drag them out of the sidebar onto the desktop or a Finder window. macOS creates a new PDF containing exactly those pages.
• Split by deletion:duplicate the file (File > Duplicate), delete the unwanted thumbnails in the copy (select, press Delete), then File > Export as PDF. Repeat per output file.
• Move pages between PDFs: open both files with thumbnails visible and drag thumbnails from one sidebar to the other.

Preview does a true structural split — text stays selectable, vectors stay vectors. Its weakness is repetition: extracting five separate ranges means five rounds of select-and-drag, and bursting a 200-page file is out of the question.

Method 3: print to PDF with a page range

Every modern OS can “print” to a PDF file, and every print dialog accepts a page range. Open the PDF (a browser tab is enough), press Ctrl+P / Cmd+P, set Pages to 12-18 or 5, and choose Save as PDF (macOS, Chrome, Firefox) or Microsoft Print to PDF (Windows) as the destination.

This works everywhere with zero new software, but it is the one method that re-renders rather than copies. The print pipeline may rasterise vector graphics, re-encode images at print resolution, flatten form fields, and drop the text layer, bookmarks, and links entirely. For a page you just need to email for reference, fine. For anything where the output must remain a first-class PDF — searchable text, fillable forms, small file size — use a structural split instead.

Method 4: qpdf or pdftk on the command line

For batch work, automation, and huge files, qpdf (open source, actively maintained, available via Homebrew, apt, and Windows binaries) is the reference tool. Its --pages option selects pages from one or more inputs:

• Extract page 5: qpdf input.pdf --pages . 5 -- page5.pdf
• Extract a range: qpdf input.pdf --pages . 12-18 -- chapter2.pdf
• Everything except pages 3–4: qpdf input.pdf --pages . 1-2,5-z -- trimmed.pdf (z means the last page)
• Burst into single pages: qpdf input.pdf --split-pages out-%d.pdf
• Split every 10 pages: qpdf input.pdf --split-pages=10 batch-%d.pdf

The older pdftk does the same jobs with pdftk input.pdf cat 12-18 output chapter2.pdf and pdftk input.pdf burst. Both tools are scriptable, so splitting 400 scanned files in a loop is a one-liner. Both run entirely offline.

What carries over — and what breaks

A structural split copies each kept page’s content stream, fonts, and images untouched, so visual fidelity is exact. The subtleties are in the document-level plumbing:

• Annotations survive. Highlights, comments, and stamps are attached to their page object and travel with it.
• Form fields mostly survive. A field on a kept page keeps working. But a calculated field that reads values from a field on a removed page silently stops computing, and splitting can duplicate field names across the output files, which confuses some viewers when the files are later re-merged.
• Links that point at removed pages break. A table-of-contents entry linking to page 40 does nothing in a file that only contains pages 1–10. External URL links keep working. Bookmarks (the outline panel) may be dropped or pruned depending on the tool.
• Digital signatures are invalidated. A signature certifies the exact bytes of the file; any split produces different bytes. The output shows the signature as broken or strips it. If a signed contract must stay verifiable, distribute the original alongside the extract.

Page numbers vs page labels

The most common splitting mistake is a numbering mismatch. A PDF has a physical page index — first page in the file is 1, second is 2 — and, optionally, page labels, display names the author assigned per section. A typical report labels its front matter i, ii, iii, iv and then restarts the body at 1. Viewers show the label, but split tools (and print dialogs, and qpdf) address the physical index.

So “extract pages 1–10” of that report, given literally, returns four roman-numeral pages plus body pages 1–6. The fix is mechanical: find the physical index of the page you want (most viewers show something like “7 of 96” alongside the label) and feed the tool physical numbers. Also note that labels themselves are document-level metadata; after a split, some tools drop them, so the extracted body may display plain 1, 2, 3 instead of its original labels.

Choosing quickly

• One-off extract or range, any OS, confidential file: in-browser split tool — local processing, no upload.
• On a Mac, no browser tab handy: Preview thumbnail drag.
• Quick reference copy, fidelity irrelevant: print-to-PDF with a page range.
• Batches, automation, gigabyte scans: qpdf or pdftk.

And the reverse trip: after extracting pieces, our PDF merge tool reassembles them, PDF compress shrinks a heavy scan before sending, and PDF to JPG turns a single extracted page into an image for a slide or chat message. All of them run in the browser, and none of them upload your document.