You scan a double-sided document on a single-sided feeder, following the sensible-looking instructions: run the stack through, flip it, run it through again. The result is a PDF with all the fronts first and all the backs afterwards — in reverse order. Or you scanned on a duplex machine and every second page is upside down. Both are standard, both are fixable in about a minute once you know the operation, and neither requires rescanning.
The three classic failure patterns
Pattern A: fronts then backs, backs reversed. You have a 20-page document scanned as two passes. Pages 1–10 of the PDF are the odd-numbered sides in order; pages 11–20 are the even-numbered sides in reverse order, because flipping the stack reverses it. This is the most common case by a wide margin.
Pattern B: fronts then backs, backs in order. Same as above, but you (or the scanner) reversed the second stack before feeding it. Pages 11–20 are the even sides in correct order.
Pattern C: every second page upside down. A true duplex scanner captured both sides in one pass, but the originals were printed with the reverse side rotated 180° — standard for calendar-flip binding, and common in some forms. Order is correct; rotation is not.
Identify which you have by looking at the last page of the PDF. In Pattern A it is the back of page 1 — the reverse of the very first sheet. In Pattern B it is the back of the last sheet. That single check tells you everything.
Fixing Pattern A: interleave with reverse
The operation you need is "collate two halves, second half reversed". Several tools name it directly.
PDFsam Basic (free, cross-platform) has an Alternate Mix module: add the file twice, set the second copy to reverse, and give it the page ranges 1–10 and 11–20. Output is correctly collated. This is the most approachable route for someone who does not want to open a terminal.
pdftk, from the command line:
pdftk A=scan.pdf shuffle A1-10 Aend-11 output fixed.pdf
shuffle interleaves the ranges one page at a time, and Aend-11 counts backwards from the end — which is exactly the reversed second half. This is a one-liner worth keeping in a notes file. See pdftk introduction.
qpdf expresses the same idea with its --collate flag:
qpdf --empty --pages scan.pdf 1-10 scan.pdf 20-11 -- fixed.pdf --collate
The descending range 20-11 handles the reversal. See qpdf introduction.
Acrobat has no single-click equivalent, which is a genuine gap. The manual route is to split the file at the halfway point, reverse the second part (Organise Pages → select all → right-click → Reverse, in versions that offer it), then insert its pages alternately — tedious enough that even Acrobat users tend to reach for PDFsam here.
Fixing Pattern B: straight interleave
Same tools, no reversal:
pdftk A=scan.pdf shuffle A1-10 A11-20 output fixed.pdf
qpdf --empty --pages scan.pdf 1-10 scan.pdf 11-20 -- fixed.pdf --collate
Fixing Pattern C: rotating alternate pages
Order is fine; you need to rotate every second page by 180°.
qpdf rotates by page range with a step, using its page-selection syntax:
qpdf in.pdf --rotate=180:2,4,6,8,10 -- out.pdf
For long documents, generate the list in a shell loop rather than typing it. pdftk uses cat with rotation suffixes: pdftk in.pdf cat 1 2south 3 4south output out.pdf — verbose, but scriptable.
Acrobat and Preview both let you select alternating thumbnails with ctrl/cmd-click and rotate the selection in one action, which is entirely reasonable for a twenty-page document and miserable for a two-hundred-page one.
General rotation ground: how to rotate a PDF page.
One caution: rotation in a PDF is a page attribute, not a transformation of the content, and some tools apply it only for display. If the rotation appears to vanish when the file is printed or converted, the fix is to flatten the rotation into the page geometry — Ghostscript reprocessing does this.
Odd page counts
If the document has an odd number of sheets, the second pass produces one fewer page (the last sheet's blank back may or may not have been captured). The interleave then goes wrong at the end, and the tools will not warn you.
Two approaches: either check and adjust the ranges manually for the last sheet, or let the scanner capture the blank back, interleave normally, and then remove the trailing blank — see how to remove blank pages from a PDF.
Verifying the fix
Do not trust the operation blindly, because an off-by-one produces a document that looks plausible and is entirely wrong.
- Check pages 1, 2, 3 — the first three should read as a continuous document, front-back-front.
- Check the last two pages. If the document ends mid-sentence or on a blank, the collation slipped.
- Check the middle. Errors that only appear halfway through come from an odd count or a double-feed.
- Check page numbers, if the original had them printed. This is the fastest verification available and the reason printed page numbers are worth having.
If the document turned out to have a page missing entirely rather than out of order, the cause is a multi-feed — two sheets pulled through together — and no amount of reordering fixes it. Rescan, and enable multi-feed detection in the scanner driver if it has it.
Preventing the whole problem
Remediation is fine; not needing it is better.
- Use a duplex scanner for double-sided documents. Almost every document scanner made in the last decade has one, and many people never turn it on because it is off by default.
- If you must do two passes, do not flip the stack end-over-end. Flipping it sideways (rotating around the long edge) keeps the order rather than reversing it, turning Pattern A into the simpler Pattern B. Which direction to flip depends on your feeder's orientation — test once with four numbered sheets and write the answer on a sticky note on the scanner.
- Check the scanner's duplex "flip on short edge / long edge" setting. This is what causes Pattern C, and it is one checkbox.
- Number your test sheets. Anyone setting up a scanning workflow should run four hand-numbered pages through first. Thirty seconds of testing prevents a two-hundred-page mistake.
- Turn on blank-page removal for mixed single- and double-sided batches.
More scanner-side settings in how to scan to PDF on Windows and how to scan to PDF on a Mac.
Doing it at scale
If this happens routinely — a records department digitising files, say — build the fix into the pipeline rather than performing it per document. A small script that takes an input file, splits at the midpoint, interleaves with reversal, then runs OCR and writes to an output folder handles a whole scanning day unattended. The pieces are all in this article; the wrapper is twenty lines. And once you are scripting, add an assertion on the page count — the single check that catches the double-feed that would otherwise reach the archive silently.
Summary
Fronts-then-reversed-backs is fixed by interleaving with a descending second range: pdftk shuffle A1-10 Aend-11, qpdf --collate with 20-11, or PDFsam's Alternate Mix. Upside-down alternate pages are a 180° rotation of the even pages. Verify at the start, middle and end afterwards — and then go and turn duplex on in the scanner so it does not happen again.
