- Steps to upload a PDF
- PDF limits at a glance
- The catch: text yes, pictures no
- Workarounds for charts and scans
- Google Docs and other formats
- Questions that get good answers
- Working with several PDFs at once
- Deleting a PDF afterwards
- Can ChatGPT read a PDF better than I can skim it?
- My verdict
- Frequently asked questions
To upload a PDF to ChatGPT, click the + in the message box, choose Add photos & files and select the PDF, or drag it into the chat, then ask your question. PDFs up to 512 MB work on Free and other plans, but outside Enterprise ChatGPT reads only the PDF's text, not the images inside it.
Uploading a PDF to ChatGPT is quick. Getting a good answer about it depends on one detail most people never hear about: on most plans ChatGPT reads the text in your PDF and skips the pictures. Charts, diagrams and scanned pages can come back as a blank. Here's how to upload, the limits, and how to work around that catch.
Steps to upload a PDF
- Open a new chat in ChatGPT.
- Click or tap the + icon in the message box.
- Choose Add photos & files and select your PDF. On a computer you can drag the file into the chat instead.
- Wait for the upload to finish.
- Type what you want, such as 'summarise this in plain English' or 'list every deadline in this contract'.
PDF limits at a glance
| Limit | What OpenAI says |
|---|---|
| File size | Up to 512 MB per document |
| Length | Capped at 2 million tokens per file |
| Free plan uploads | 3 file uploads per day |
| Other plans | Up to 80 files every 3 hours, lower at peak times |
| Files per project on Free | Up to 5 |
Two small but annoying points: failed uploads can count toward your limit, and ChatGPT doesn't tell you how much of your allowance is left. If you're on Free, get the file right before you upload it.
The catch: text yes, pictures no
OpenAI's file uploads FAQ explains that support for images inside documents depends on your plan. ChatGPT Enterprise can do visual retrieval on PDFs. Every other plan uses text-based retrieval, which pulls out the digital text and ignores the embedded images.
In practice, that means a chart in an annual report, a diagram in a manual or a table saved as a picture may not be read at all. And if your PDF is a scan of paper pages, there may be little or no digital text for ChatGPT to extract in the first place.
Workarounds for charts and scans
- Screenshot the chart or page and upload it as an image. ChatGPT reads images directly, which avoids the text-only route.
- For a scanned document, run it through a text-recognition tool first so the PDF contains real text, then upload it.
- Copy any important table into a spreadsheet and upload that as an Excel or CSV file instead.
- Ask ChatGPT what it can see. If it summarises only the words around a chart, you know it skipped the chart.
Google Docs and other formats
You can't upload a .gdoc file directly. OpenAI says to export it as a PDF or Word document first, then upload the exported file. Word (DOCX) and plain text (TXT) files work too, and often give cleaner results than a heavily designed PDF.
Questions that get good answers
OpenAI's own examples are a good guide: summarise a research paper in simpler terms, compare two documents, find every reference to a topic, extract quotes, or pull out the headings and lists. I'd add: always ask for the page or section an answer came from, so you can check it.
- 'Summarise this report in five bullet points for a busy business owner.'
- 'List every date, deadline and notice period in this contract.'
- 'Compare these two proposals and show where they differ.'
- 'Quote the exact sentence that supports each point.'
Working with several PDFs at once
You can attach more than one PDF to a chat and ask ChatGPT to compare them, which OpenAI lists as one of the things uploads are good for. Each file counts toward your upload limit, so on Free you'll only manage three in a day. If you keep coming back to the same set of documents, put them in a project instead of uploading them again in every chat. On Free, a project holds up to 5 files, and other plans allow more.
Name your files clearly before uploading, such as 'Supplier A quote' and 'Supplier B quote', then refer to them by name in your question. It makes answers easier to follow and cuts down on mix-ups between documents.
Deleting a PDF afterwards
Here's one people miss. Deleting the chat doesn't delete the file if it was saved to your Library. OpenAI says to open Library, select the file and delete it separately. If the PDF contained anything sensitive, do both. My guide to how private ChatGPT is covers the wider data settings.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
Can ChatGPT read a PDF better than I can skim it?
For finding a clause or summarising a long report, often yes. For anything you'll act on, like a contract term or a number in a table, check the original page. ChatGPT is quick, not infallible, and it can't read what it can't see. If you're curious how it decides what to say, my guide to where ChatGPT gets its information explains the basics.
My verdict
Upload the PDF with the + button, ask specific questions and ask where each answer came from. If your document is full of charts or is a scan, upload screenshots or a text-recognised version instead, because on every plan but Enterprise, ChatGPT is reading the words, not the pictures.
Frequently asked questions
How do I upload a PDF to ChatGPT?
Click the + in the message box, choose Add photos & files and select the PDF, or drag it into the chat.
Can I upload PDFs to ChatGPT for free?
Yes. Document uploads work on the Free plan, which OpenAI limits to 3 file uploads per day.
What is the PDF size limit in ChatGPT?
OpenAI says documents can be up to 512 MB per file and are capped at 2 million tokens.
Can ChatGPT read charts and images in a PDF?
Only ChatGPT Enterprise supports visual retrieval for PDFs. Other plans extract the digital text and skip embedded images.
Does deleting the chat delete my uploaded PDF?
Not if it's saved in Library. OpenAI says to delete the Library copy separately.


