A file won't open. How to work out why
"The file is corrupt" is what software says when it means "this is not what I expected". Most of the time the file is fine and something else is wrong.
The extension is a label, not the format
Renaming a file changes nothing about its contents. Every real format begins with a signature — a few bytes that identify it regardless of what the name claims — and that is what applications actually read.
Mislabelling happens constantly and innocently: a website serves a PNG from a URL ending in .jpg, an export names everything .xls when it wrote CSV, someone renames an extension hoping to get past an upload filter. The file is intact; only the label lies. Checking the real signature turns an unopenable file into a working one in about ten seconds.
Truncated downloads
A download that was interrupted leaves a file that looks entirely normal — a sensible name, a plausible size, no error anywhere — and is simply missing its end. PDFs are unusually good at revealing this, because the index that tells a reader where the pages are sits at the very end of the file. Lose the last few kilobytes and every page becomes unreachable at once.
The tell is a file noticeably smaller than the original, or one that opens and shows the first part before failing. Re-download it before spending time on repair; a partial file cannot be recovered, because the missing bytes were never written.
Encryption, which does not announce itself
A password-protected PDF often fails with a generic error rather than a password prompt, especially in a browser viewer or a mobile preview. There is a second, more confusing case: a document with no open password but with an owner password restricting editing. It displays perfectly and then refuses every attempt to modify it, which reads exactly like corruption.
The right file, the wrong application
HEIC on Windows, WebP in older software, a modern DOCX in a very old Word, a CSV using semicolons opened by a spreadsheet expecting commas — in each case both the file and the program are working correctly and are simply not talking about the same thing. The fix is conversion, not repair.
The semicolon case is worth singling out, because it does not fail loudly: the file opens, and every row lands in a single column. Nothing is broken. The delimiter is regional.
Actually corrupt
Rarer than the reputation suggests. Real corruption comes from failing storage, a bad transfer over an unreliable connection, or software that crashed mid-write. Text-based formats — JSON, CSV, XML — can often be repaired, because the damage is usually structural and visible: a trailing comma, an unclosed bracket, a stray byte-order mark at the start of the file that makes the first key unreadable.
Binary formats are less forgiving. If a JPG's data is damaged, the part after the damage is genuinely gone. Diagnose before you spend an afternoon on repair — knowing a file is unrecoverable early is worth more than a slow discovery.