Key takeaways
- OpenAI says ChatGPT reads the text in a PDF, so scanned PDFs made only of images need OCR before they can be analyzed.
- To upload, open the tools menu in the message box, choose Add photos & files, pick the PDF and send it with a clear question.
- OpenAI lists a 512MB limit per file and 2 million tokens per document, and free accounts have a daily upload cap.
In this story 7 sections
You have a 40 page contract, a research paper or a product manual, and you want answers without reading every line. Learning how to upload a PDF to ChatGPT takes about a minute, and this guide is for anyone using ChatGPT in a browser, the desktop app or the phone app. The upload itself is quick. The part that takes a little more thought is checking that your file will actually be readable and asking questions that get you specific, useful answers.
Check your PDF before uploading
- Check that the PDF has real text. OpenAI's File uploads FAQ says ChatGPT analyzes the text in a PDF, and PDFs that contain only images (no text at all) are not supported. A quick test: open the file and try to highlight a sentence. If you can select words, there is a text layer.
- Know the size limits. The same FAQ lists a hard cap of 512MB per file and 2 million tokens per text or document file.
- Know your plan's upload allowance. OpenAI notes that free users can upload a limited number of files per day, paid plans get far more, and limits can tighten during busy hours. If you hit a cap often, our ChatGPT Plus vs free comparison explains what changes when you pay.
- Remove anything you would not want stored. Uploaded files stay with the chat until you delete it, so redact account numbers, medical details or client names first if you can.
Uploading the PDF on the web or desktop app
OpenAI's guide to working with files in ChatGPT describes the same basic flow for every type of document.
- Open ChatGPT and start a new chat. A fresh chat keeps earlier conversations from muddying the answers.
- Open the tools menu in the message box at the bottom of the screen.
- Choose Add photos & files. (OpenAI's guide shows this label in the menu, alongside options like Web search and Deep research.)
- Pick your PDF from your computer. You can also drag the file straight into the chat window.
- Wait for the file name to appear as an attachment above the message box, then type your question in the same message and send it.
Sending the file together with a clear question works better than sending the file alone and waiting. ChatGPT reads the document in light of what you asked, so the first reply is more focused.
Uploading from an iPhone or Android phone
The FAQ confirms that file uploads work in the iOS and Android apps as well as on the web. On a phone the steps are nearly the same:
- Open the ChatGPT app and start a new chat.
- Open the tools menu next to the message box and choose the option to add files.
- Browse to the PDF in your phone's file manager. Email attachments usually need to be saved to your files first.
- Add your question and send.
If you mostly use ChatGPT on Apple devices, our guide on using ChatGPT on iPhone covers the app's other features.
Adding a PDF from a cloud drive
If your document lives in a cloud service rather than on your device, ChatGPT can sometimes reach it through an app connection. OpenAI's guide says to go to Settings, then Apps, find the service you want, and complete the sign in and permissions screens. Connected apps then show up in the tools menu during a chat, or you can bring one up by typing @ or /. In business and Enterprise workspaces, an admin decides which apps are available, so you may not see every option.
Getting answers you can trust
Uploading is the easy half. The questions you ask decide whether the result is useful.
- Start with an overview. Ask for a short summary and a list of the main sections. This shows you quickly whether ChatGPT read the whole file or only part of it.
- Ask for specifics with locations. "List every deadline in this lease with the page or section it appears in" gives you something you can check. OpenAI's own example prompt is to extract key dates and owners from a PDF into a simple table.
- Check the important claims. Open the PDF and confirm any number, date or clause you plan to rely on. ChatGPT can misread or merge details, especially in long documents.
- Narrow the scope for long files. If a reply feels vague, point it at one chapter or section at a time.
- Compare documents in one chat. You can upload two versions of a policy and ask what changed between them.
Good prompting habits carry over from regular chats. Our guide to writing better AI prompts has more patterns that work well with documents.
Charts and pictures inside the PDF
According to the File uploads FAQ, when a PDF mixes text and images, ChatGPT analyzes the text, and images embedded in documents are not read as pictures. In practice this means a chart's caption and surrounding paragraphs are available, but the bars themselves may not be. If a chart matters, take a screenshot of that page and upload it as an image in the same chat, then ask about it directly.
When the PDF upload does not work
Most failures come down to a few causes.
- The file is a scan. Scanned pages are pictures of text, not text. Run the file through an OCR tool (many PDF editors and scanner apps include one), save a new copy and upload that.
- You hit an upload limit. Free accounts have a daily cap, and OpenAI says limits may be lowered during peak hours. Wait and try again later, or use a paid plan.
- The file is too large. Split a huge PDF into smaller parts, such as one file per chapter, and upload the part you need.
- The PDF is password protected. Remove the password in your PDF reader (you need to know it) and save an unlocked copy.
- Storage is full. OpenAI caps storage at 10GB per user, and you will see an error when you reach it. Deleting old chats with large files frees space; the FAQ says files from deleted chats are removed within 30 days.
- ChatGPT itself is having trouble. If uploads fail for every file, the problem may be on OpenAI's side. Our ChatGPT not working guide walks through the checks to run.
OpenAI's ChatGPT capabilities overview explains that when you upload a document such as a PDF, ChatGPT uses tools to extract the text and find the relevant parts. That is why a clean, text based file gives better answers than a blurry scan. For more guides like this one, browse our AI section.
One last habit worth keeping: before you close the chat, ask ChatGPT to list any parts of the document it could not read. If it flags pages or tables, you know exactly where to look yourself.
Sources
Share this story
Spotted an error? Tell us


