How to extract pages from a PDF — pick them visually, keep the original

Pages 12 to 18 of a ninety-page contract, the one appendix that matters in a scanned manual, the map pages of a travel guide — pulling a subset out of a PDF is one of the most common document tasks there is, and most tools make it harder than it should be. You type “2, 5-7” into a box, hope the count was right, and download the result to find out. This guide covers the extraction that keeps your file on your machine and your eyes on the pages themselves, then the decisions most guides skip: extraction versus deletion, extraction versus splitting, and how ordering works in the new document.

Extract pages in your browser, step by step

The direct path is the Extract Pages tool on this site. It runs entirely in your browser — the document is never uploaded.

1. Open the extraction tool and drop your PDF anywhere on the page, click to browse, or paste it from the clipboard.
2. Every page appears as a thumbnail of the real document. Tick the pages that will make up your new file — a live counter tracks the selection as you go.
3. Tap any page to open a full-size preview. Verifying that page 14 really is the appendix takes one click, not one download and scroll.
4. Reorder if the job calls for it. The extracted pages keep the on-screen order of the cards, so drag them — or use the arrows — until the sequence reads the way you want. Page 9 may legitimately precede page 2.
5. Click Extract pages. A new PDF is assembled on your device from your selection. The original file is never modified, moved, or re-saved.

Pages are copied byte-for-byte rather than re-rendered: text stays selectable, links still work, scanned pages keep their resolution.

Extraction and deletion are two faces of one operation

Here is the model that makes the choice trivial. Extraction keeps what you checked and drops the rest; deletion drops what you checked and keeps the rest. Same operation, opposite reading of the selection — extracting pages 2, 5 and 9 produces exactly the document that deleting every other page would.

So the rule is arithmetic, not preference: count the smaller side. Pulling 4 pages out of a 60-page report is 4 checks on the extractor; removing 4 pages from that report is 4 checks on the deletion tool. Either tool can reach either result in theory; only one of them lets you finish before lunch.

And because both tools write a new file and leave the source untouched, the decision is reversible. A wrong call costs a minute — reopen the original tomorrow and run the other reading.

Pointing at pages beats typing ranges

Range syntax was designed for command lines, not for people. It asks you to know page numbers before you can verify them, which is why the miscounted extraction is such a classic failure: the wrong page ships, and nobody notices until the recipient replies.

A thumbnail grid inverts the workflow — you see the pages first, then choose. Scanned documents benefit most: when every page looks like a gray rectangle from 2009, counting is guesswork and seeing is verification. The built-in zoom preview closes the remaining gap, because a scan’s printed page numbers routinely disagree with the PDF’s own order — front matter, unnumbered inserts and plates shift everything by a few pages.

Scattered selections cost nothing extra: pages 3, 7 and 11 in one pass, kept in exactly that order if the job wants it.

Extract, delete, or split?

Splitting deserves a seat at this table because the overlap is larger than people expect. Split carves a document into ranges in a single run — the right tool when the whole file must be distributed, every chapter as its own document. Extraction shines when one subset becomes one document and the rest is irrelevant.

The overlap case is concrete: pages 1–3 as one file and 8–9 as another. Extract twice, or split once with the ranges “1-3, 8-9” — one operation either way. A nuance tips repeated work toward splitting: the extraction selection resets after each run. For a one-off it is irrelevant; if you rebuild the same subsets every week, entering both ranges in a single split is the shorter path.

Where extraction fits in a bigger workflow

Two chains cover most real projects. First, locked sources: an encrypted PDF must be unlocked before anything else — the browser cannot display pages it cannot read, so extract from the unlocked copy. Second, assembly: pull the section that matters from each source document, then merge the pieces into one file. The pages arrive in the order you set at extraction time, so the merged result needs no second pass.

As for scale, browser tools handle documents of a few hundred pages comfortably. A 2,000-page scanned archive on an aging machine is where a desktop application with direct memory access starts to earn its installation — though even there, extracting a handful of pages is usually faster than installing anything.