Two very different complaints turn out to have the same underlying mechanism. A page that opens on its side, and a page with margins so wide the content is lost in the middle — both are fixed by changing how the page is presented, without altering a single mark that was drawn on it.
Understanding that is what stops you from making the file worse while fixing it.
Rotation is a number, not a redraw
Every page in a PDF carries a dictionary of properties, and one of them is /Rotate. Its value is 0, 90, 180 or 270, and it instructs the reader to display the page turned by that much before showing it.
Nothing else changes. The text is drawn at the same coordinates in the same fonts. The images sit in the same places with the same compression. The reader applies a transformation at display time.
This is why rotation done properly is completely lossless. Rotate PDF turns selected pages by changing the page dictionary, so text stays selectable and images keep their original quality.
It is also why a badly built rotation tool is so destructive. The naive implementation renders each page to a bitmap, rotates the bitmap, and writes it back as an image. The page now looks right and has quietly become a photograph of itself — unsearchable, uncopyable, several times larger, and permanently degraded.
The test takes two seconds: open the rotated file and try to select a line of text. If you cannot, the page was flattened to an image and you should discard that copy and start again from the original.
Why the page was sideways to begin with
Worth knowing, because it tells you how much to rotate:
- A scanner fed the sheet in landscape. Usually every page is wrong in the same direction — rotate all of them.
- A phone photo carried an EXIF orientation flag that the converter ignored. Often affects some pages and not others.
- The document genuinely mixes orientations — a report with a wide table on page 12. Here the sideways page is correct and rotating it is the mistake. Check whether it reads properly when turned before assuming it is broken.
Rotate only the pages that are wrong. Applying a blanket rotation to a mixed document swaps one problem for another.
Cropping is also just a boundary
A PDF describes several rectangles for each page. The two that matter here:
- MediaBox — the physical sheet the page was designed for
- CropBox — the region a reader should actually display
Cropping sets the CropBox. The reader shows that rectangle and hides the rest. Crop PDF also sets the MediaBox to the same rectangle, so viewers and printers agree on the new page size. It changes only the page boxes, without re-encoding the objects already inside the document — so, exactly like rotation, the content keeps its original quality. Auto crop PDF does the same, with a box it measures from where the content sits on each page.
The part that catches people out
Cropped content is hidden, not deleted.
Everything outside the crop box is still in the file. Enlarging the page box again brings it straight back, and doing so requires no particular expertise.
This matters because cropping looks like a way to remove something — a header with an internal reference, a footer with a file path, a margin note. It is not. Anyone who receives the document can recover it.
If content must genuinely be gone, it has to be removed rather than hidden. Delete PDF pages writes a new document whose page list leaves the page out, but if a page you keep links to it, or the pages share images and other resources, parts of the deleted page can still be copied into the file. For a page that must not survive, rebuild the document from its source without it. For part of a page, cropping is the wrong instrument entirely; use Redact PDF, which rebuilds each page you mark as an image. Pages you leave unmarked are copied as they were, so the same caveat applies to them.
Printing a cropped file
The second surprise: a crop can look right on screen and not print.
Different consumers read different boxes. Viewers honour the CropBox. Some printers and print workflows use the MediaBox, so a crop that moves only the CropBox can print the full original sheet, including the area you thought you had trimmed away. Crop PDF and Auto crop PDF move the MediaBox too, so the page is the new size everywhere.
If the file is destined for print, test it early on the actual printer rather than trusting the on-screen preview. A printer set to fit the page to the paper will enlarge a cropped page.
Cropping a scan, specifically
This is the case where cropping genuinely earns its place. A scanned page often carries a black border, the edge of the platen, or the shadow of the facing page. Trimming to the sheet makes it read as a document rather than a photograph of one.
Crop generously. A tight crop that clips a descender or the edge of a signature is worse than a loose one, and unlike the hidden-content problem, this damage is visible to your reader.
A sensible order of operations
When a scan needs several corrections, sequence matters:
- Rotate first, so every page is the right way up.
- Crop second, once you can see what you are trimming.
- Then run OCR PDF if the text needs to be searchable — recognition is markedly more accurate on a straight, tightly cropped page than on a rotated one with a black border.
- Compress last, if size is an issue, because compression settings depend on what the pages finally contain.
Doing OCR before rotation is a common and expensive mistake: the recognition runs against sideways text, produces nonsense, and has to be redone.
Before you send it
- Open the saved copy, not the one still in the editor.
- Select some text to confirm nothing was flattened to an image.
- Scroll every page if you applied a blanket rotation — the page that was already correct is now wrong.
- Print one page if the file is for print, before committing to the whole run.