Why a Photo Looks Right on Your Phone and Sideways Everywhere Else

You take a photo holding your phone sideways for a wider shot. It looks completely normal in your phone's gallery - upright, correctly framed. Upload the exact same file to some other site or app, and it shows up rotated 90 degrees, as if the phone had been held the "wrong" way after all. Nothing about the image data changed between those two views. What changed is whether the software displaying it bothered to read one small piece of information stored alongside the pixels.

The camera never actually rotates the pixels

When you hold a phone sideways to shoot a photo, the camera sensor still captures the image in whatever physical orientation the sensor itself sits in - it doesn't rotate the raw pixel data to match how you were holding the phone. Instead, the phone's motion sensor detects the orientation at the moment of capture and writes that as a small tag inside the photo's EXIF metadata: "Orientation: 6" (or one of several other standard values), meaning "this sensor-orientation image should be displayed rotated 90 degrees clockwise to look correct." The actual pixel grid stored in the file is unrotated sensor data; the correct on-screen orientation lives entirely in that one tag.

Most software reads the tag. Some doesn't.

Modern operating systems, photo gallery apps, and current web browsers almost universally read and apply this EXIF orientation tag automatically and correctly - which is why the photo looks right in your phone's gallery, and usually right in a browser too. The mismatch happens with software that decodes the raw pixel grid without checking the tag: an older image-processing tool, a script that reads pixels directly without going through a full image-decoding library, some social platforms' upload pipelines at certain points in their processing, or a canvas-based re-encode performed by a tool that doesn't specifically account for orientation. The result is the sensor's raw, untransformed orientation showing through - sideways or upside down relative to what you actually saw when you took the photo.

Same pixels, one small tag decides everything

The eight possible values, and why you don't need to memorize them

The EXIF orientation tag isn't limited to one rotated case - it has eight standard values covering every combination of 0/90/180/270-degree rotation and horizontal mirroring, since a camera sensor can theoretically be read in any of those eight configurations depending on how the device was held and how its sensor is physically mounted. You don't need to memorize what each numbered value means; the point worth taking away is that this is a small, fixed, well-documented set of possibilities, not something arbitrary or unpredictable - which is exactly why software that's built to check for it can always resolve it correctly, and software that skips the check always gets it wrong in one of these eight specific, predictable ways.

Why this is worth fixing at the file level, not per platform

Chasing this issue platform by platform - re-uploading, hoping the next site handles it better - is unreliable, since it depends entirely on decisions made by software you don't control and can't predict in advance. The dependable fix is to correct the orientation once, in the file itself, so the pixels genuinely reflect the correct upright orientation rather than relying on an EXIF tag some future piece of software might or might not honor. A rotated-and-flattened file looks correct everywhere, permanently, because the fix no longer depends on any reader correctly interpreting a metadata tag - the pixels themselves are already right.

What actually needs to happen

Rotating the image so its pixel data matches its intended orientation - not just relying on the tag - fixes this permanently for that file, regardless of what displays it next. The rotate/flip image tool does exactly this: it reads the file the same standards-compliant way a modern browser does (correctly applying EXIF orientation automatically during decode), then any rotation or flip you apply redraws the actual pixel grid, producing a file that displays correctly everywhere from then on, tag or no tag.

This isn't limited to camera photos

The same underlying mismatch shows up beyond phone photography too - a scanned document, an image exported from certain editing software, or a photo passed through several tools in sequence can all end up with an orientation tag that doesn't match how a particular piece of software chooses to interpret it. Anywhere an image genuinely looks correct in one place and wrong in another, with no visible corruption or quality loss, the EXIF orientation mismatch is the first thing worth checking - it's a far more common explanation than anything actually being broken about the file.

For the actual mechanics of mirroring or rotating a photo, see how to mirror an image online. And for correcting a whole folder of photos that came out this way at once, see fixing a batch of sideways photos.