Your file is processed entirely in your browser and never uploaded: the PDF is opened in memory on your own device, its pages are copied into a new document in the sequence you specify, and the rearranged file is offered back as a download. Nothing is sent to a server and the tool works offline once loaded. Pages are copied as objects, not re-rendered, so reordering changes the order and nothing else — every page keeps its content, text layer, size and rotation exactly.
The sequence is a list of original page numbers in their new order. Writing “3, 1, 2” for the opening of a document puts the old page 3 first, followed by the old pages 1 and 2. A sequence that lists every original page number exactly once is a pure rearrangement: a 5-page file given “3, 1, 2, 5, 4” comes out with 5 pages, the third and the swapped pair moved. Counting the entries before you run it is the check — the output page count always equals the number of entries in your sequence.
Because the sequence is just a list, the same mechanism covers two neighbouring jobs with open eyes: an entry repeated twice copies that page twice (a deliberate duplicate, for example a form page that must appear in two places), and an original page left out of the sequence does not appear in the output at all — which is deletion by omission, the mirror of Delete PDF Pages. Typical honest uses are fixing a scanner that stacked pages in the wrong order, moving an appendix before the conclusion, or putting a misplaced page back where it belongs.