Split PDF

You need something that can copy pages into a fresh document rather than an editor that only hides them. Here, drop the PDF in and the page count appears beside the filename, type what you want in the Pages box — single numbers, ranges written with a hyphen, separated by commas — then press Extract Pages and take the download. Everything you type is measured against the length of the document before any file is built, so a range that runs past the end, or one written backwards, stops with a message naming the part that is wrong. That is deliberate: the alternative is a download that is quietly missing pages you asked for.

Because they are counting different things. Printed numbering usually starts after a cover, and long documents often give front matter roman numerals before the arabic numbers begin, so the page marked 1 can be the third sheet in the file. Some viewers show that printed label rather than the position, which makes the gap easy to miss. Positions are what get used here: page 1 is the first sheet, whatever it says on it. The reliable check is to scroll to the page you want and read where it sits out of the total your viewer reports, then use that number rather than the one on the paper.

Usually, but rarely in proportion to the pages you dropped. A PDF stores fonts, embedded colour profiles and images once and points at them from wherever they are used, so the saving depends on what the pages you kept still need. Take three text pages out of a 200-page report and the result can be tiny; take the three pages that carry all the photographs and you keep most of the weight. Nothing is re-compressed during the copy either, so image quality and image size come across untouched. If a size limit is the real problem, extract first and then run the result through a compressor.

Not here. The file is read before anything else happens, and an encrypted document fails at that step — you get a message saying it could not be read and may be encrypted or corrupted, and no page count appears. There are two kinds of PDF password, one that stops the file opening at all and one that claims to restrict printing or editing, and neither is handled. The fix is to open the document in a reader where you know the password, save an unprotected copy, and split that. It is also worth not typing a document password into a random website in order to have the protection stripped.

No. The pdf-lib library is pulled into the page and the whole job — reading the document, counting its pages, copying the ones you named, saving the result — happens in JavaScript on your own machine. The download is a temporary link held by the tab, and each new file or new run releases the previous one, so an old link cannot be downloaded by mistake after you have changed your mind. If a file arrives from another tool on the site it is handed across inside the same tab and the page says so on screen. That matters more than usual here, given that the documents people cut up tend to be contracts, bank statements and letters from a doctor.

The pages will. What sits above them will not. The output begins as an empty document and receives copied pages, so the bookmark tree, the title and author metadata, and the structure that makes form fields fillable are not carried over — the widgets may still be drawn on the page without the machinery behind them. Internal links are the subtle one: a cross-reference pointing at page 140 has nothing to point at once you have kept only pages 4 to 9, so it either dies or lands somewhere unintended. If the outline of a long document is part of its value, keep the original alongside the extract rather than replacing it.

Not by extracting them. Whatever order you type the numbers in, the list is sorted into ascending order before the copy runs, so an extract always comes back in the document's own sequence. To move a block, take each piece as its own extract and then join them in the order you want with Merge PDFs: pages 5 to 8 as one file, pages 1 to 4 as another, then merge with the second half first. That works for one or two moves and gets tedious quickly. For a real reshuffle, a desktop editor with a thumbnail view you can drag will cost you less time.

If the original had selectable text, yes. Pages are copied across as they are rather than redrawn or re-encoded, so characters remain characters and a viewer's find command works on the extract exactly as it did on the source. If searching the result turns up nothing, it would have turned up nothing in the original too — that document is a scan, meaning a photograph of text rather than text, and no amount of splitting will change it. Reading those words as characters needs OCR, which is a separate step. Splitting neither adds a text layer nor destroys one.