How to analyze documents with ChatGPT: a practical guide

If you want to know how to analyze documents with ChatGPT, the short version is this: you upload the file, ask a clear question, and ChatGPT reads the text and answers. The real work is choosing the right question, knowing the format limits, and checking the answer against the original document before you act on it.

How do you upload a file to ChatGPT?

Uploading a file takes three steps. Open a chat, click the paperclip icon next to the message box, then choose the file on your device. According to OpenAI's own guide, this works with CSV, XLSX, PDF, DOCX, JPEG, PNG, TXT and more (OpenAI Academy). You can also drag the file straight into the chat window, and the same upload icon appears in the mobile app.

Once the file is attached, ask a direct question instead of a vague one. "Summarize the three main points of this report" works better than "tell me about this file." If some of your documents are in Azerbaijani and you are new to the tool itself, our guide to using ChatGPT in Azerbaijani covers the basics first.

What file formats and limits should you know?

Each format fits a different kind of document. The table below lists the common ones.

A larger file means more text for the model to process; in technical terms, text is split into small units called tokens. A very long document uses a lot of tokens, which can make answers slower or cut off mid-way. If your file is very large, split it or point your question at one specific section instead of the whole thing (OpenAI Academy). You can also upload more than one file in the same message, which helps when comparing an old and a new version of a contract. For spreadsheet-specific wording, our ChatGPT prompts for Excel page has ready templates.

How does ChatGPT handle complex PDF documents?

PDFs sometimes have a very complex structure, and the text inside them can come out in the wrong order, especially in scanned or multi-column pages (OpenAI Developer Community). This happens because a text-based PDF and a scanned image saved as a PDF are read very differently by the model. If the text in your file cannot be selected or copied, you are working with a scanned document that needs OCR, which stands for optical character recognition, the process of turning an image of text into real text.

In practice, some users convert the scanned pages into images first and then rely on the model's image-reading ability instead (OpenAI Developer Community). Our guide to analyzing images with ChatGPT walks through that approach. The simple rule: if the output looks jumbled or incomplete, open the source file yourself and check the order of headings and tables.

What tasks can you ask ChatGPT to do with a document?

A specific prompt gets a specific answer. OpenAI's own examples by role show the range well:

• Finance: turn a quarterly statement into a board-ready summary
• Marketing: review a campaign brief or summarize a competitor research PDF
• Sales: polish a proposal draft saved as DOCX or PDF before sending it
• HR: pull common themes out of CSV or XLSX employee survey answers
• Operations: check uploaded checklists and audit reports for compliance gaps

(Source: OpenAI Academy). The same source lists sample prompts worth copying almost word for word: "Summarize the main findings in this report," "Visualize this sales data by region," "Check this contract for overly complex language." If contract language is your actual concern, our guide to reviewing contracts with AI goes deeper into that specific case.

How do you manage several documents at once?

Sometimes one file is not enough. A contract draft, an earlier version, and a set of notes might all live in separate files, and re-uploading all of them every time you start a new chat gets tedious fast.

ChatGPT Projects lets you keep a group of files and instructions together in one place, so the context, meaning the background information the model keeps in mind, does not need to be re-explained in every conversation. This is especially useful for legal or finance teams that work from the same template documents again and again, though someone still needs to track when the files themselves go out of date.

Contracts, reports and spreadsheets: practical examples

Picture a small bakery owner who uploads a month of sales data in XLSX and asks which weekday consistently underperforms and why. This is a hypothetical example, but it matches the kind of task OpenAI lists under "visualize this sales data by region" (OpenAI Academy).

Another hypothetical case: a freelance bookkeeper uploads a client's annual PDF report and asks for the three main findings in plain language. Or a one-person HR consultancy uploads a CSV of survey answers and asks which complaint shows up most often. In every case, the useful habit is the same: ask one clear question at a time, and treat the first answer as a draft, not a final number.

How do you check the result before you trust it?

ChatGPT can sometimes produce an answer that sounds confident but is factually wrong, a behavior known as hallucination (Wikipedia, ChatGPT). In document work, this usually shows up as a misquoted figure, a wrong date, or a misremembered name. Do not pass a financial number, a contract term, or a client's name into a decision until you have checked it against the real document yourself.

Three mistakes worth avoiding

• Forwarding the answer to a client or manager without reading it first
• Uploading one huge file and expecting the same prompt to cover every question in it
• Uploading confidential client data before checking your company's data policy

Start with a small, low-risk document, compare the answer against the source yourself, and only move to larger or more sensitive files once that check feels routine.