2026-07-25

Why your PDF rotation isn't saving — how page rotation actually works

"PDF rotation not saving" is one of the most common PDF complaints there is — search it and you'll find guide after guide promising to fix it. Most of the time the rotation did happen. The file just doesn't behave the way people expect afterward, because a PDF page's rotation is a single stored number with its own particular rules, and most rotate buttons don't explain any of them.

What actually changes when you rotate a page

A PDF page carries one entry in its own page dictionary called /Rotate. It only takes four values: 0, 90, 180, or 270 degrees, always clockwise. That's the whole mechanism. Rotating a page doesn't touch the content stream — the actual instructions for the text, images, and vector lines on that page — at all. It changes a single "display this many degrees rotated" instruction and leaves everything else byte-for-byte identical. That's also why rotating a 40-page PDF is close to instant and doesn't grow the file: there's no image to re-encode, just an integer to update on however many pages you touched.

Two different things both get called "rotate"

Search "PDF rotation not saving" and the complaint is always the same shape: rotate a page, close the file, reopen it, and it's back to how it started. Many free PDF readers offer two controls that look identical but do different things. One is a view rotation — a display preference for that reading session only, stored in the reader's own settings, never written into the file itself. The other is a page rotation that rewrites the page's actual /Rotate value and saves a new file, which every other app, every printer, and the next person who opens it will also see. This is an established enough distinction that PDF vendors document it by name — Bluebeam has a support page titled, plainly, "Rotate View vs. Rotate Pages." If a rotation doesn't stick, it's almost always the first kind: nothing was ever written to the file.

Some pages arrive pre-rotated before you touch anything

Not every sideways page got that way from someone clicking rotate. Scanners and phone-scanning apps often capture a page in whatever orientation the sensor happened to be in, then just set /Rotate to correct the display, instead of re-encoding the image the right way up. That keeps the output small, but it means a scanned PDF can arrive with real, non-zero rotation already baked into individual pages — sometimes a different value per page, if sheets were fed through the scanner in alternating directions. Opening a document like that and finding page 4 sideways doesn't mean anyone rotated it wrong. It means the scanner already wrote a rotation value, and anything you do next happens on top of that value, not on a blank slate.

Rotation adds up, it doesn't reset

That "on top of" part matters, because rotating a page doesn't set its orientation to a fixed value — it adds your chosen amount to whatever rotation the page already had. A page that already carries 90° and gets rotated another 90° by a tool ends up at 180°, not 90°. This is exactly what NearPDF's own rotate implementation does: page.setRotation(degrees((page.getRotation().angle + rot) % 360)) reads the page's current angle and adds the requested rotation, wrapping back to 0 past 360. Four clicks of a 90° rotate button always return a page to where it started — but it also means that if a page still looks wrong after one click, the fix is usually another click, not a broken button.

Why the same file can look right in one app and wrong in another

Even a correctly-written /Rotate value isn't a guarantee every piece of software downstream honors it the same way. Print pipelines are a known trouble spot: Mozilla's own bug tracker has a filed report of its built-in PDF viewer failing to print rotated pages correctly — the on-screen rotation was right, the printed page wasn't. That's not something a rotate tool can prevent from its side. Writing a standards-correct /Rotate value into the file is the entire job of a page rotation tool; whether the next viewer, printer driver, or converter in the chain reads that value correctly afterward is a property of that other software, not of the file a rotate tool hands off.

How NearPDF applies this

NearPDF's rotate tool works on individual pages in the workspace grid: each click of a page's rotate control adds 90° to that page, shown live as a preview so you can see it before saving anything. Saving combines that accumulated per-page rotation with whatever /Rotate value the page already had and writes the result into the output PDF via pdf-lib — a real, permanent change to the file, not a view setting — entirely inside a Web Worker on your own device, with nothing uploaded. Reorder, delete, and rotate all operate on the same page list before you save once, so a sideways scanned page, a page you want gone, and a page out of order can all be fixed in a single pass instead of three separate tools and three separate exports.