Fixing a document you cannot re-export
Pages end up in the wrong state constantly, and usually for reasons outside your control. A scanner fed a batch sideways. Two passes of a double-sided original came out interleaved wrongly. A document arrived with a duplicate cover, a blank separator page, or an appendix somebody wanted moved to the front.
When the source file exists, fixing it there is always better. Often it does not: the document came from a client, a portal, a system nobody maintains, or a scanner. Rearranging the PDF itself is then the only route, and it is a genuinely lossless one — pages are copied whole into a new order, so nothing is re-rendered and nothing degrades.
Three operations cover nearly everything
Rotate for pages that came out of a feeder the wrong way up. Move for order. Delete for the blanks, the duplicates and the pages that should not be in the copy you are about to send.
The [msm_tool slug="pdf-page-organizer"] reads the document, lists its pages, and lets you work through them before saving a rebuilt file. The original on your disk is untouched — you get a new document, which is the right default when the operation is destructive.
Rotation is a property, not a picture
Worth understanding because it explains a recurring confusion. A rotated page in a PDF is not a rotated image; it is the same page content with an instruction saying which way up to display it. Nothing is re-encoded and no quality is lost, however many times you turn it.
It also means the underlying text keeps its original orientation. A sideways scan turned upright displays correctly and, if it has a text layer, still selects and searches correctly, because the text was never sideways in the first place — only the page was.
The practical note: rotate in ninety-degree steps and check the result rather than assuming. A page that looks upside down may be at 180 or may be at 90 twice, and the difference is invisible until you save.
Deleting is not redacting
The important warning in this whole area, and it has embarrassed serious organisations.
Removing a page removes that page. It does not remove anything from the pages that remain, it does not touch the document's metadata, and it is not a way of taking sensitive material out of a file while keeping the rest. Those are separate operations with separate tools.
Where a page has to go because of what is on it, delete it, then check the result properly — search the output for a phrase that was on the deleted page, and look at the document properties. The page is gone; anything it shared with the rest of the file may not be.
Interleaving a double-sided scan
The most common real repair, and worth spelling out because people do it by hand and get it wrong.
A double-sided original scanned in two passes gives you all the fronts in order, then all the backs — and the backs are usually in reverse, because the stack was flipped as a block. The correct document is front 1, back 1, front 2, back 2, and getting there means reversing the second half before interleaving it.
Check the middle of the result rather than the beginning. The first two pages will look right under almost any wrong arrangement; the failure shows up halfway through.
What a rebuild does not carry
Rearranging produces a new document, and some document-level structure does not survive the move. Bookmarks pointing at pages by position become wrong or are dropped. Internal cross-reference links pointing at a deleted page go dead. A form split or reordered may confuse fields that referenced each other.
For a scan or a straightforward report, none of this applies — there was nothing to lose. For a long structured document with navigation, expect to rebuild the outline afterwards, and factor that in before deciding whether the source file is really unavailable.
Work in one pass
Do the rotations, the moves and the deletions together, then save once. Saving between each operation means several rebuilds, several files, and a real chance of continuing from the wrong one.
Then read the output end to end at thumbnail size. It takes under a minute for most documents, and it catches the sideways page you missed, the duplicate you deleted the wrong copy of, and the page that ended up one position from where you meant it — all of which are much cheaper to find now than after the document has gone out.