Splitting takes one PDF and produces several, either by page ranges you specify or by breaking out every page as its own file. Ranges are written the way you would say them aloud — 1-3, 7, 10-12 — and each comma-separated group becomes one output document, so that example yields three files rather than six. The pages are copied into new documents rather than re-rendered, which means text stays selectable, vector artwork stays sharp and the visual result is identical to the original. Everything runs in your browser: the file is never uploaded, so there is no size limit and no waiting behind other people's jobs. Multiple outputs are bundled into a single ZIP for download. The usual reason to split is that someone needs one section of a long document and should not receive the rest.
How to write the ranges
A range is a group of pages that will end up in one file. Write single pages as a number, spans with a hyphen, and separate the groups with commas: 1-3, 7, 10-12 produces three documents — one holding pages 1 to 3, one holding page 7, one holding pages 10 to 12.
Page numbers here mean physical positions in the file, counting from 1, which is not always what is printed on the page. A report with unnumbered covers and roman-numeral front matter will have a printed page 1 several positions in. Check against the page thumbnails rather than the printed numbers.
Splitting against extracting
The two overlap and the difference is what you get back. Splitting is for dividing a document into several parts you keep — chapters, one file per invoice, a copy per recipient. Extracting is for pulling a few pages out and ignoring the rest.
In practice, if your ranges cover the whole document you are splitting; if they cover a fraction of it you are extracting. Both produce the same kind of output, so the choice is about which mental model matches what you are trying to achieve rather than a technical distinction.
What comes across into each part
Each output is a real PDF with the original page content, links and annotations intact. Because pages are copied at the object level, a split file opens and prints identically to the corresponding pages of the original.
Links that pointed to pages now in a different output file cannot resolve and are dropped, as are bookmarks referring to pages outside the range. Form fields spanning the split are the one case worth checking by eye before sending anything on: a field that belonged to a page in another part will not come with it.