Splitting a PDF sounds trivial until you try to send somebody a 90-page manual, one clause of a contract, or everything except the annex that contains somebody else's medical data. The operation itself is a page-level rewrite - build a new page tree that references the pages you want, in the order you want them - and every serious tool does it, locally or online. Knowing what it can and cannot do matters more than usual, because the most common disappointment is people trying to extract a rectangle. These tools cut at page boundaries. They will not hand you the top half of page 12 as a new file, because producing that means rendering the page and cropping the result, which turns text into an image and silently makes the document unsearchable and inaccessible to screen readers. The other recurring surprise is that splitting is not always the privacy control people assume it is: metadata, hidden layers and embedded attachments can carry over into the extracted file. This page links out; the notes below are about the sharp edges.
Splitting happens on page boundaries
Every mode these tools offer is a range operation. Split by range produces one output per range you type, extract every page produces one file per page, and extract a selection produces a single file made of the pages you ticked. All three are the same underlying operation: write a new page tree pointing at the chosen page objects, and copy across whatever those pages need. What none of them can do is a partial page. If you need one table, one signature block or one diagram out of the middle of a page, you are in crop-and-render territory: the result will be an image, the text will no longer be selectable or searchable, and if you need the text back you will also need optical character recognition. Two things that live outside the page tree are worth knowing about: files embedded inside the PDF as attachments can be dropped when a page is extracted, and layers used for technical drawings or maps are optional content that some tools flatten.
Ranges, single pages and the parts that disappear
Decide what you actually want before you pick a mode. Extract every page is the right answer when the plan is to send a batch of separate images or documents, and it produces the most files and the most wasted structure. Split by range is right for chapter-per-file, and it keeps the sequence under your control. Extract a selection is right when you just need to hand over a few pages. In all cases, expect the document outline to be lost or truncated to the part that made it into the file, and expect the bookmarks, comments and form fields on the extracted pages to be unreliable. If the pages you are extracting carry the digital signature, be aware the signature will not be carried over as a valid signature for the new file.
Large files and the honest cost of a link-out tool
Splitting a large scanned document means uploading it, waiting for the queue, and downloading the results. Server-side tools set their own limits: a per-file size cap, a per-task page count, and a daily quota on the free tier, and those change over time, so check the current note rather than trusting an old blog post. A 300-megabyte scan of a bound manual will usually hit one of those limits. When that happens the local route is genuinely better: qpdf and pdftk can split a file of any size that fits on your disk, with no upload, no queue and no waiting, and they are free. The web tool earns its place when the file is small, the job is one-off, and you would rather not install anything.
Frequently asked questions
Can I extract just one paragraph from the middle of a page?
No, and that is a structural limit rather than a missing feature. These tools cut at page boundaries. Anything smaller means rendering the page and cropping it, which produces a flattened image with no selectable or searchable text. If you need the text back afterwards, that is a separate OCR step, and the accuracy on scans will not be perfect.
Are the extracted pages uploaded anywhere?
Yes. The service linked from this page processes files on its own servers, so its privacy policy and retention rules apply to what you send. If the document contains personal or confidential material, split it locally with a command line tool such as qpdf or pdftk, which never touches the network.
Is splitting free and are there limits?
The free tier covers normal documents and imposes per-file, per-task and per-day limits that the provider can change at any time. Read the current limits on their page before a large job, and expect a wait in the queue for bigger files. Local tools such as qpdf have no such limits, only your disk space.
Do bookmarks and links still work after splitting?
Partially at best. The document outline is usually rebuilt only for the pages that stayed in each output, so entries pointing to removed pages disappear, and internal links can turn into dead references. After splitting, open one of the outputs and check the navigation panel before you send it on.