5 free ways to get more file uploads out of ChatGPT

First, know which cap you hit

The fixes are different, so this is worth thirty seconds.

Storage is the one people get stuck on, because it does not clear with time. It is cumulative across every chat, Project and custom GPT you have, and a file from months ago still counts. Check Settings, then Storage.

1. Paste instead of uploading

The one that actually gets you unstuck, and most people never think of it.

The cap counts uploads. Pasted text is a message, not an upload. So when you are completely out of uploads, you can still open the document, copy the text, and paste it in. It works immediately, with no waiting and no payment.

One condition, and it is new. Since June 2026, pasting more than 10,000 characters in one go makes ChatGPT convert the paste into an attachment instead of putting it in the text field, and an attachment spends an upload. Keep each paste under that and it stays a message. If a paste does convert, "Show in text field" on that paste puts the text back.

1. Open the document on your computer

   Any editor or reader. For a PDF this needs real text rather than a scan, which you can test by trying to select a sentence.

2. Copy the text

   Select all, copy.

3. Paste it in, in parts if it is long

   Each part under 10,000 characters. Tell ChatGPT first that a document is coming in several parts and not to answer until the last one arrives.

More manual than uploading. Also completely unblocked by the limit that is stopping you.

2. Stop retrying failed uploads

OpenAI's documentation confirms failed attempts can count toward the rate cap.

On Free, with 3 file uploads per day, this is brutal: one file that fails three times has spent your entire day and uploaded nothing.

If an upload fails, find out why before trying again. Too large fails identically every time. Too long uploads but gets truncated. A scan uploads fine and comes back empty. None of those improve with repetition.

3. Bundle files before uploading

The cap counts files, not bytes, so six documents concatenated into one text file is one upload rather than six.

This works especially well for code. One file with a header line before each source file keeps everything ChatGPT needs to understand the structure, and costs a single slot:

for f in src/auth/*.ts; do echo "=== $f ==="; cat "$f"; done > bundle.txt

4. Send only what you need

You rarely need a whole document read. You need the section your question is about.

One chapter is one upload, and it usually produces a better answer than the full document, because there is less irrelevant material competing for attention. Two wins for the price of one slot.

5. Clear storage if that is the wall

If Settings then Storage shows you near 25 GB, waiting will never help.

Deleting the chat no longer does it. Files saved to your Library stay there when the chat holding them is deleted, which is a change from how this used to work and the reason the old advice to clear old conversations stopped having any effect. Delete the files themselves, from Settings then Storage or from the Library, and remember the cap is shared with your Projects and custom GPT knowledge, so look there too.

Worth doing periodically rather than only when blocked, since nothing prompts you.

What does not work

Zipping files together. ChatGPT does not unpack archives, so a zip is one upload that yields nothing. Bundling into a text file achieves what people are hoping for from zipping.

Renaming extensions. Files are inspected by content, not by name.

A second account. The caps are per account so it would work, and it likely breaches OpenAI's terms. Risking an account you depend on to save a few uploads is a poor trade when pasting achieves the same thing.

Upgrading, if your problem is length. Paid plans lift the daily count but not the 2 million tokens cap that truncates long documents. If a single long file is your problem, paying changes nothing.