ChatGPT will not accept that file type. Convert it, do not rename it.

The message arrives immediately, usually as the file lands in the composer rather than after a progress bar. That timing is the useful part. Nothing was uploaded, no ceiling was reached, and no parser tried and gave up. ChatGPT looked at what kind of file you handed it and stopped there.

OpenAI publishes no list of accepted extensions. The entire official answer is one sentence: "All common file extensions for text files, spreadsheets, presentations, and documents." Four families, with the word common doing quiet work in the middle of them. Things outside those families, and a fair number of uncommon things inside them, get turned away at the door.

Which refusal you actually got

Three different problems produce three messages that people describe with the same words.

Only the first row is about format. If your file was accepted and the trouble started afterwards, you hit a ceiling or an empty extraction, and neither of those is fixed by changing the extension.

The file name is not what decides

There are two gates, and they are easy to confuse.

The first is the file picker. Your browser is told which types the page will take, so unsupported files show up greyed out. That is a convenience filter working from the extension and the type your operating system reports, not an inspection, and on most systems you can override it by switching the dialog to show all files. Dragging a file straight onto the page skips it too. Both are how people end up looking at the refusal in the first place.

The second gate is the parser that has to open the bytes and pull text out of them. That one does not care what the file is called. A file named report.docx that is not a Word document inside is still not a Word document, and the code that expects Word structure will not find it.

So a matching extension is necessary and not sufficient. Getting past the picker means you were allowed to try, nothing more.

Renaming makes it worse, with one exception

Changing the extension changes nothing about the contents. In the good case you get a second failure a few seconds later, having spent an upload attempt for nothing. In the worse case something is extracted, it is not your document, and there is no signal telling you that. Confidently wrong output from garbled input is harder to catch than an error message.

The exception is when the new name is true. Plenty of files are plain text already and just happen to carry an extension nobody registered: .log, .yml, .ini, .tex, .srt, .conf. Saving one of those as .txt is not a disguise, it is a correction. The bytes were always text.

What to convert to

Convert toward what the file actually contains rather than toward what feels closest.

Two things follow from that table. Archives are not a format problem you can convert your way out of, because the content is other files and ChatGPT does not unpack them. And design files have picture content, so the honest conversion is a picture, capped at 20 MB each, not a document.

For anything that is genuinely words, plain text is the most reliable target of the lot. There is no structure to misread and nothing to strip.

When there is no clean conversion

Print to PDF is the general escape hatch. Anything that can be printed can become a PDF, and PDF is accepted. One caveat: printing preserves whatever was there. If the source was already a picture of text, the PDF is a picture of text too, and you trade this problem for an empty answer instead.

Copying the text into the message box is the other fallback, and it sidesteps file rules entirely because nothing is uploaded. It is governed by message length rather than by format, so it suits a clause, a page, or a short email, and not a report.

For audio and video, neither is among the four families OpenAI lists, so there is no conversion that makes the media file itself acceptable. Get a transcript and upload that.

Do not keep retrying the same file

OpenAI's own FAQ says failed upload attempts can count toward the rolling cap of 80 files every 3 hours. It does not say which failures count, so treat repeated attempts as expensive rather than harmless. On the Free plan, where the allowance is 3 file uploads per day, four hopeful retries with a format that was never going to work can leave you unable to upload the converted version once you have it.

The practical decision

Read the timing of the failure first. Refused as you attach it means format, and format has exactly one fix, which is conversion. Work out what the file really holds, export to the closest accepted form of that, and upload once. Leave the extension alone unless the new one tells the truth.