ChatGPT Says “An Unknown Error Occurred”? Check the File
ChatGPT won't tell you what's wrong with your upload. Drop the PDF here and see for yourself — password, size, page count, missing text, broken file — then split or re-save it in one click.
Drop the PDF ChatGPT refused
It is checked in your browser. Nothing is uploaded.
Opening…
- Free
- No sign-up
- No ads
- Runs in your browser, nothing is uploaded
What to try, in order
Check the file above
You get a list: does it open at all, is it password-protected, how big is it, how many pages, and do the pages actually contain text.
Fix what the list flags
Long or heavy file → split it into parts and upload them one at a time. Odd internal structure → save a clean copy. A scan with no text → turn it into text instead.
If everything passes, it isn't your file
Re-save a clean copy anyway, then check OpenAI's status page, try another browser or the phone app, and retry later.
Why ChatGPT shows this error
“An unknown error occurred” is the message ChatGPT shows when reading the file didn't finish and it has nothing more specific to say. OpenAI has not published what triggers it, and people hit it on PDFs, spreadsheets, slides and even plain text files, so there is no single cause to point at.
What you can do is rule out the file. A password-protected PDF is refused before anything is read. A scan contains no text at all. A very long or heavy document has more that can go wrong on the way up, and splitting it is the quickest thing to try. And a PDF written by an unusual exporter sometimes opens in your reader but trips up other software — copying every page into a fresh file gets rid of that.
This page checks those things on your own machine and gives you the two fixes that don't need any other software.
What each check means
| Check | Why it matters | What to do |
|---|---|---|
| Opens as a normal PDF | If a standard reader can't open it, nothing else will either | Export a fresh PDF from the original document |
| Password | Protected files are refused before the text is read | Open it with the password, export a copy without one |
| File size | Heavier files have more to go wrong on the way up | Split it and upload the parts |
| Page count | Very long documents are the other common case | Split into parts of 25 pages |
| Text inside the pages | A scan has no text, so ChatGPT finds nothing to read | Turn it into text and paste that |
| File name | Unusual characters in the name break some uploads | Rename it to plain letters, numbers and dashes |
What it won't do
- It can't tell you what ChatGPT thought. The error is generic and nothing about the failure is visible from outside, so this is a checklist of the things you can rule out, not a diagnosis.
- It can't remove a password. If the file needs one to open, this page can't read it either.
- It can't shrink images. “Save a clean copy” rewrites the file structure; it doesn't re-compress photos, so a scan-heavy PDF stays about the same size. Split that one instead.
- It doesn't talk to ChatGPT. You upload the fixed file yourself; nothing is sent to OpenAI from this page.
FAQ
What does “An unknown error occurred” mean in ChatGPT?
It's a generic message: ChatGPT couldn't finish reading the file and doesn't say why. OpenAI hasn't published what triggers it, so the practical approach is to rule out what you can check on your own file — a password, a large size, a very long document, a scan with no text, or a file a PDF reader can't open — and then re-save or split it.
Is my file uploaded anywhere?
No. The PDF is opened, checked, split and re-saved in your browser. The PDF reader library is downloaded to your browser, but your file never leaves the page.
My file passes every check. What now?
Then the file probably isn't the problem. Save a clean copy anyway, check OpenAI's status page, try another browser or the phone app, and wait a while before retrying.
Does splitting lose anything?
Pages are copied as they are, so text, images and formatting stay. What you lose is anything that spans the whole document: bookmarks, form fields and the table of contents.
Can it remove a password?
No. If a PDF needs a password to open, this page can't read it either, and nothing here tries to break it. Open it with the password in a PDF reader and export a copy without one.
What about Word, Excel or images?
This check is built for PDFs. For other files the generic fixes still apply: rename it to plain letters and numbers, re-save it from the app that made it, and split very large files.