How to upload a large PDF to ChatGPT: 4 ways that work

PDFs are the format people most often need ChatGPT to read, and the one that fails in the most confusing ways. A rejected upload at least tells you something went wrong. The other two failure modes look like success.

First, work out which failure you have

Three minutes here saves you from fixing the wrong problem.

1. Did the upload finish?

   If it refused outright with a size error, you have a size problem. That is the rare one and the easy one.

2. Ask about the last page

   Upload succeeded? Ask something that can only be answered from the final few pages. "What is the last section called?" works. If it cannot tell you, the document was truncated.

3. Ask it to quote a sentence

   Ask for a direct quote from page two. If it returns nothing at all, or says the document appears to have no readable text, you have a scan.

If it is a scan

This is the most common cause of "ChatGPT says my PDF is empty", and it is not a bug.

Except on Enterprise, ChatGPT extracts the digital text from a file and discards the images. A scanned PDF is a photograph of a page. There is no digital text in it, so the extraction returns nothing.

The fix is OCR, which converts the picture of the words back into words:

• Mac: open in Preview, export as PDF, and recent versions handle the text layer automatically.
• Adobe Acrobat: Scan and OCR, then Recognise Text.
• Free web tools: plenty exist. Do not use one for anything confidential.

Test it worked before uploading again: open the PDF and try to select a sentence with your cursor. If you can highlight the words, there is real text in there now.

If it is too long

This is the most common failure on real, text-based PDFs, and the one with no error message. The four routes below are the PDF-specific version of the general methods for bypassing the ChatGPT file upload limit.

The relevant ceiling is 2 million tokens per file, not the 512 MB size limit. A three-hundred-page report can be a few megabytes and still be well past what fits. When that happens the file uploads normally and ChatGPT works from as much as it could hold.

Four ways out, roughly in order of how much effort they cost.

Send only the part you need

The one people skip, and often the best answer. You rarely need all three hundred pages read. You need the section your question is about.

Extract those pages into their own PDF, or copy that text into the chat directly. You will usually get a better answer than from the full document, because there is less competing for attention.

Convert it to text first

If the PDF is text-heavy, exporting to plain text strips out fonts, page geometry, and embedded images, and what is left is often a fraction of the size. It does nothing for the token count if the words themselves are the bulk, but it frequently helps more than expected.

Split it and send the parts in order

The reliable one. Cut the document into pieces and send them one after another.

The instruction matters more than the splitting. Send this first:

I am going to send a long document in several parts. Do not summarise or answer anything until I tell you the last part has arrived. Just reply "received" after each one.

Without it, ChatGPT answers part one as though it were the whole document and then keeps revising as more arrives. The result is weighted heavily toward the opening, which is worse than useless because it looks complete.

Use something built for very long documents

Past a certain size, chunking stops being sensible. Even sent in order, the earliest parts fall out of memory before the last ones land. For a book or a full archive, a retrieval tool that searches the document and pulls out relevant passages will beat any amount of splitting.

Knowing when to stop fighting the tool is worth more than another workaround.

If it is genuinely too large

Rare, but real. The ceiling is 512 MB per file, and a PDF gets there by containing high-resolution images.

Compressing the PDF usually solves it outright, and since images get discarded anyway on every plan below Enterprise, you lose nothing ChatGPT was going to read. Most PDF tools have a compress or reduce-file-size option. Export to text works too, for the same reason.

What does not work

• Zipping it. ChatGPT does not open archives. You have added a wrapper, not solved anything.
• Renaming the extension. Files are read by their contents, not their name.
• Upgrading to Plus. Some caps move. The length limit that truncates long documents does not.
• Uploading it again. If it failed for size or length, it fails the same way every time.

The short version

Scan? Run OCR, then upload.

Too long? Send the section you need, or split it and tell ChatGPT to wait for all the parts.

Actually too large? Compress it. You are only losing images it was going to discard anyway.