A 60-page annual report arrives and you need pages 4, 17 and 52 for a slide deck. A scanned contract has one signature page the lawyer wants on its own. A textbook chapter runs from page 112 to page 140 and the rest is dead weight on your tablet. In each case the job is the same: make a new PDF that contains only the pages you choose, and leave the source file exactly as it was.
This guide covers how page extraction works inside a PDF, how page-list syntax behaves at the edges, what survives the trip into a new file and what does not, and when a different operation (split, delete or reorder) is the better fit. The examples use the Extract PDF Pages tool on Toolz, which runs entirely in the browser.
TL;DR: Extracting PDF pages copies the pages you select into a brand-new document and leaves the original untouched. Type page numbers and ranges such as
1, 5, 9or10-12, or click numbered tiles, and download<name>_extracted.pdf. The output always follows the original page order, lists each page once, and keeps each page's text, fonts and images without re-encoding. Bookmarks, internal page links, fillable form fields and document metadata do not carry over. The free PDF page extractor is client-side, no signup, and works offline once loaded.
What does extracting pages from a PDF do?
Extracting pages creates a new PDF document and copies a chosen set of pages from the source into it. Nothing is removed from the original file; you get a second file containing the subset.
Under the hood a PDF keeps its pages in a structure called the page tree, and each page is an object that points at its own content stream plus the resources it uses: fonts, images, color spaces. The format is specified in ISO 32000-2. Extraction walks that tree, copies the selected page objects and every resource they reference into a fresh document, and writes the new file. Because the page content is copied rather than redrawn, text stays selectable, vector graphics stay vector, and images are not recompressed.
The Toolz tool does this with pdf-lib, a JavaScript PDF library, running inside your browser tab. The file is read into memory, the new document is built there, and the result is offered as a download. No copy of the PDF is sent to a server.
Extract vs split vs delete vs organize: which one do you need?
These four operations overlap, and picking the wrong one usually costs an extra step. The difference is how many files come out and which pages you name.
| Operation | You name | Output | Best when |
|---|---|---|---|
| Extract pages | The pages to keep | One new PDF | You want a few scattered pages from a long file |
| Split PDF | Where to cut | Several PDFs | You want a long file broken into consecutive parts |
| Delete pages | The pages to remove | One new PDF | You want most of the file minus a few pages |
| Organize PDF | A new order | One new PDF | You want the same pages in a different sequence |
A rule of thumb: count the pages you would have to type. From a 40-page contract, keeping 38 pages means naming 38 in extract or 2 in delete, so delete PDF pages is faster. Keeping 3 of 40 is the reverse. When the goal is several files rather than one, the guide to splitting a PDF covers range-based cutting.
How to extract specific pages from a PDF, step by step
- Load the file. Drag a PDF onto the drop zone or click to browse. The tool shows the file name, size and page count, then one numbered tile per page.
- Select pages. Type numbers and ranges into the page field, click tiles, or combine both. Select All fills every page and Select None clears the selection.
- Check the selection. A counter reads, for example, "3 of 12 pages selected", and a "Pages to extract" line lists the chosen pages in order.
- Download. The button changes to "Extract 3 Pages". Clicking it saves a new file named after the original with
_extractedadded, soreport.pdfbecomesreport_extracted.pdf.
The typed field and the tiles stay in sync. Clicking tiles 4, 5 and 6 writes 4-6 into the field for you, and if the field already says 1-3, 8, clicking tile 4 rewrites it as 1-4, 8. Tiles show page numbers, not thumbnails of the content, so if you need to find a page by what is on it, open the PDF in a viewer first and note the numbers.
How does page-range syntax behave? Before and after examples
Page numbers count from 1, matching what a PDF viewer shows. The input is a comma-separated list where each item is a single page or a range written start-end. The examples below all use a 12-page file called report.pdf.
| You type | Pages in the output, in order | Why |
|---|---|---|
1, 5, 9 |
1, 5, 9 | Three single pages |
9, 1, 5 |
1, 5, 9 | Output always follows the original order |
1, 5-7, 12 |
1, 5, 6, 7, 12 | Singles and ranges can be mixed |
3, 3, 7 |
3, 7 | A page is included once, however often it is typed |
4-6, 5-8 |
4, 5, 6, 7, 8 | Overlapping ranges merge |
10-15 |
10, 11, 12 | A range past the last page is clipped |
0, 13, 2 |
2 | Numbers outside 1 to 12 are ignored |
7-3 |
nothing | A backwards range selects no pages |
5- |
nothing | An open-ended range is not read |
1-12 |
1 to 12 | The whole document, which is what Select All writes |
Here is the before-and-after view for the first example, laid out as page sequences:
Source report.pdf 1 2 3 4 5 6 7 8 9 10 11 12
Input 1, 5, 9
Output report_extracted.pdf 1 5 9
(new page numbers) 1 2 3
The extracted file has three pages, and they are numbered 1, 2 and 3 inside the new document. Any printed page numbers in the content itself still say 1, 5 and 9, because extraction copies the page as it was drawn.
Three of those rows are easy to trip on. A backwards range such as 7-3 and an open range such as 5- select nothing and show no warning, and an out-of-range number is silently skipped. The counter is the check: if you expected 5 pages and it says 3, look at the field before clicking Extract.
Can you extract pages in a different order, or the same page twice?
No to both, by design. The selection is a set of page numbers, and the output is built from that set in ascending order. 9, 1, 5 and 1, 5, 9 produce identical files, and 3, 3, 7 produces a two-page file.
When you need a custom order, do it in two passes. Extract the pages first, then open the result in the PDF page organizer and drag them into sequence. The organize PDF guide walks through reordering in detail.
When you need a page to appear twice, for instance a cover sheet at both the front and back of a packet, extract that page on its own and then combine the files with Merge PDF:
Pass 1 extract 1 -> cover.pdf [1]
Pass 2 extract 2-11 -> body.pdf [2..11]
Pass 3 merge cover.pdf + body.pdf + cover.pdf
Result packet.pdf [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 1]
What survives extraction, and what gets lost?
The page itself travels whole. Things that belong to the document rather than to a single page stay behind. This split follows directly from how ISO 32000 structures a file: some objects hang off a page, others hang off the document catalog.
| Item | Carried over? | Reason |
|---|---|---|
| Text, vector graphics, images | Yes, unchanged | Part of the page's content stream and resources |
| Fonts | Yes | Referenced by the page, copied with it |
| Page size and rotation | Yes | Page attributes |
| Links to websites | Yes | Link annotations are stored on the page |
| Links that jump to another page of the document | No, they stop working | They point at page objects in the source file |
| Bookmarks (outline) | No | The outline tree belongs to the document catalog |
| Fillable form fields | No, only their appearance | The interactive form is registered at document level |
| Title, author, keywords | No | Document metadata is not copied |
| Password protection | Cannot load | Encrypted files are refused |
The form-field row surprises people most. A filled-in field still looks filled in, because its visual appearance is drawn on the page, but you can no longer click into it and type. If you need a fillable subset, keep the original and fill it there.
Does extracting pages reduce quality?
No. Pages are copied at the object level, so there is no rasterizing, no image recompression and no font substitution. An extracted page renders exactly like the same page in the source.
File size is a different matter. A one-page extract can still be large, because the page brings along every resource it uses. A single page with a full-page photograph and an embedded font can easily weigh a megabyte on its own. If the result is too big to email, run it through the PDF compressor; that step does change images, so keep the uncompressed extract if quality matters.
Is it safe to extract pages from a confidential PDF?
Yes, as far as the tool is concerned. The PDF is opened in browser memory, the new document is assembled there, and the download is generated locally. Nothing is uploaded, and once the page has loaded the tool keeps working with the network off. That makes it reasonable for contracts, medical letters or HR documents that should not pass through a third-party server.
Two practical notes. First, the extracted file is a normal, unencrypted PDF, so treat it with the same care as the source. Second, because metadata is not copied, the new file starts with empty title and author fields, which is often a small privacy gain when sharing an excerpt.
What errors can you hit, and how do you fix them?
| Message or symptom | Cause | Fix |
|---|---|---|
| "Please select a PDF file" | The file is not recognized as a PDF | Check the extension and that the file is a real PDF, not an image or Word file |
| "Failed to load PDF file" | The PDF is password-protected, damaged or truncated | Remove the password in the app that created it, or re-download the file |
| Extract button disabled | No pages selected | Type at least one valid page or click a tile |
| Fewer pages than expected | Backwards range, open range, or out-of-range numbers | Read the counter and the "Pages to extract" line |
| "Failed to extract pages" | The new document could not be built, most often memory pressure on a very large file | Close other tabs, or extract in smaller batches |
| Internal links do nothing | They pointed at pages in the source file | Expected; links to websites still work |
There is no fixed upload cap because nothing is uploaded. The practical limit is your device's memory, which matters on phones with scanned PDFs of several hundred megabytes.
Alternatives to an online page extractor
A browser tool is convenient when you are on a machine you do not control or need no installation. Other routes fit other situations.
On a Mac, Preview can do it by hand: open the sidebar thumbnails, select pages, and drag them to the desktop or into a new document. It is quick for a handful of pages and slow for dozens.
Print to PDF with a custom page range works in most browsers and operating systems, but it regenerates the file through the print pipeline, which can flatten form fields and change how text and images are stored. Extraction copies pages as they are.
On the command line, qpdf does the same object-level copy and is the right choice for scripting many files:
qpdf --empty --pages report.pdf 1,5,9 -- report_extracted.pdf
qpdf --empty --pages report.pdf 10-12 -- appendix.pdf
qpdf also accepts ranges like z for the last page and can repeat pages, which the browser tool deliberately does not.
A worked example: building a slide-deck appendix
Say a 60-page annual report needs three exhibits for a board slide deck: the revenue table on page 4, the regional breakdown on pages 17 and 18, and the auditor's note on page 52.
- Load
annual-report.pdf. The tool shows 60 tiles. - Type
4, 17-18, 52. The counter reads "4 of 60 pages selected" and the list shows4, 17, 18, 52. - Click Extract 4 Pages.
annual-report_extracted.pdfdownloads with four pages, numbered 1 to 4 inside the new file. - The auditor's note should come first in the deck, so open the extract in the organizer and move page 4 (formerly 52) to the front.
Final order in the appendix: 52, 4, 17, 18 from the source, and the 60-page original on your disk is untouched. For the rest of the PDF set, including rotate, merge and compress, see the PDF tools hub, or go straight to the page extractor and try it with your own file.
