A scanned document rarely comes out uniformly wrong. One page is sideways because the feeder grabbed it crooked, the next three are fine, then a whole section is upside down because someone reloaded the stack the wrong way round. The temptation is to rotate the entire file and hope for the best, which fixes one group of pages and breaks all the others. The correct approach is slower to describe but faster in practice: fix each group of pages by its own range, and leave the good pages alone.

Rotate PDF is built for exactly that job. It rotates all pages, or only a chosen range, by 90, 180, or 270 degrees, in your browser. The angle you choose adds to any rotation the page already has, which is a detail that trips people up and is worth understanding before you start. This post explains the rule, then walks through a fictional mixed-orientation scan fixed range by range.

Why scans come out sideways

Scanners and phone cameras do not know which way up a page is. They record pixels in the orientation the page was fed or photographed, and the PDF stores that orientation as a rotation value. When a stack of pages is fed by hand, small differences accumulate: a page placed landscape in a portrait stack, a section fed back to front, a single page turned to fit a wide table. Automatic feeders compound the problem, because a page that skews or flips in the feeder is scanned exactly as it arrived, and the operator often does not notice until the whole batch is done.

There is also a human factor. When you scan a thick report, you naturally rotate wide pages — tables, diagrams, fold-outs — to fit them through the scanner, intending to turn them back later. "Later" arrives when a reader opens the file and finds a sideways table in the middle of portrait text. None of this is a fault in the document. It is the normal friction of turning paper into a file, and it is fixable without rescanning anything.

The rule that matters: rotation adds up

Here is the rule to internalise before touching any page. A PDF page has an existing rotation, even if that rotation is zero. When you apply a rotation, the new angle is added to the existing one, and the result wraps around at 360 degrees. Rotate a page that is already at 90 degrees by another 90, and it ends up at 180. Rotate it by 270, and it ends up at 0 — upright again, because 90 plus 270 is 360, which is the same as no rotation.

This matters in two practical ways. First, you cannot assume a page needs "90 degrees" just because it looks sideways: a page that looks sideways clockwise needs 90 degrees one way, and a page that looks sideways anticlockwise needs 270 (or, equivalently, 90 the other way). Look at the page, decide which way it needs to turn, and choose accordingly. Second, if you rotate the same range twice by accident, you have not undone anything — you have added rotation twice. If a range comes out wrong, work out its current total and apply the correction that brings it to upright, rather than repeating the original rotation and hoping.

A related point: rotating a page changes only its orientation. It does not re-read the text, straighten skewed lines, or make scanned text searchable. A page that was scanned upside down and then rotated upright will read correctly to a human eye, but for text-recognition purposes the underlying image was still captured upside down — a distinction that matters if you later run text recognition on the file, and one to keep in mind rather than discover afterwards.

Worked example: a fictional mixed scan

Imagine a fictional twelve-page site report, scanned by a fictional engineer, Tom Hughes, for a fictional firm at example.org. Tom pages through the scan and notes the orientations:

  • Pages 1 to 4: upright, fine as they are.
  • Pages 5 to 7: sideways clockwise — the section was fed landscape.
  • Page 8: upright.
  • Pages 9 to 11: upside down — the stack was reloaded the wrong way.
  • Page 12: upright.

Tom's job is three separate rotations, not one. He rotates pages 5 to 7 by 90 degrees in the direction that brings them upright, and pages 9 to 11 by 180 degrees. Pages 1 to 4, 8, and 12 are left untouched. He does this in Rotate PDF by selecting each range in turn, downloading the result, and opening it to check every page — not just the ones he changed, but the untouched ones too, to confirm they stayed untouched.

Suppose Tom had instead rotated all twelve pages by 90 degrees. Pages 5 to 7 would be fixed, but pages 1 to 4, 8, and 12 would now be sideways, and pages 9 to 11 would be sideways in a different direction. He would then face a harder repair than the one he started with, because the pages would no longer fall into clean ranges. The range-by-range approach is not fussiness. It is what keeps a simple job simple.

If the scan also has pages in the wrong order, that is a separate job. Organize PDF reorders pages visually from thumbnails, and Crop PDF trims uneven scanned borders — both useful after rotation, neither a substitute for it. And if a rotated scan still needs its text made searchable, OCR PDF is the tool for that step, run after the pages are upright.

A checking routine that catches slips

After rotating, open the result and page through it from start to finish. This takes a minute for a short document and is worth it at any length, because rotation slips are easy to make and hard to spot from a page count or a file size — neither changes when you rotate. Check in particular the boundaries between ranges: the last page of one range and the first page of the next are where a mistyped range shows up, either as a page rotated that should not have been, or a page missed that should have been turned.

Also check the document in the reader your audience will use, if you know it. Rotation is stored in the file, so it should display consistently, but a quick check in the actual reading environment removes doubt. If a page still looks wrong after rotation, re-examine it before rotating again: work out its current orientation first, then apply the single correction that brings it upright, rather than stacking rotations on top of each other.

The guide to merging and splitting PDFs covers the neighbouring jobs of joining and dividing documents, which often follow a rotation fix when a scan is being prepared for distribution.

Honest limitations

Rotate PDF only changes orientation. It does not fix scanned upside-down text for text-recognition purposes: a rotated image still carries the pixels it was scanned with, and recognition tools read the image, not the display angle, unless they are run after the pages are upright and can handle the result. Rotation also does not straighten skew, de-warp curved pages, or improve image quality — a crooked scan stays crooked, just upright. Rotation is stored as an angle added to the page's existing rotation, so repeated careless rotations accumulate rather than cancel out; always determine a page's current orientation before rotating it again. Finally, rotation does not change page size or content, so a landscape table rotated into a portrait document will still be a landscape-shaped page — it will simply display the right way up.

FAQs

How do I rotate pages in a PDF?

Open the file in Rotate PDF, choose all pages or a specific range, pick 90, 180, or 270 degrees in the direction the pages need to turn, and download the result. Then page through the result to check every page, especially the boundaries between ranges.

Can I rotate only the sideways pages and leave the rest?

Yes. That is the main reason to use ranges: select only the pages that are wrong, rotate them, and leave the upright pages untouched. Rotating the whole file to fix a few pages will break the pages that were already correct.

Why do my scanned pages come out sideways?

Because scanners and cameras record pages in the orientation they were fed or photographed, without knowing which way up the content is. A page fed landscape, or a stack reloaded the wrong way, is scanned exactly as it arrived.

Does rotating a PDF change its quality?

No. Rotation changes only the orientation stored for each page; it does not re-encode images or alter text. A rotated page keeps the same quality — and the same skew or blur — it had before.

human review pending client designation