Jump to a section
The short answer

ChatGPT can open CSV and Excel files, write and run code to analyse them, and draw charts. Free users can do this with usage limits. Remove account numbers and names before uploading, ask specific questions, and spot-check a couple of totals yourself: it can make mistakes, and you can ask it to show the steps it used.

How it works

OpenAI says ChatGPT can analyse spreadsheet formats such as Excel and CSV, create charts, and run Python calculations for some analysis tasks. Not every answer necessarily runs code; review the analysis details when it does.

Make the file easy to inspect: one row per record, clear column headings and one related dataset at a time. Limits depend on your plan, model, file and remaining allowance.

Everyday uses

If your bank, energy supplier or app lets you download your data as a spreadsheet or CSV, you can ask questions about it. A few ideas:

  • Spending: export a few months of transactions from your banking app and ask where the money goes
  • Energy: download your usage data and ask which months and times cost most
  • A club or group: who's paid their subs, and who hasn't
  • Fitness: export your runs or steps and chart the trend
  • A home project: tidy up a messy list of quotes, then compare them side by side

Useful AI prompts for everyday life: copy and adapt

Clean it up before you upload

Bank exports often include account numbers, payee details and references. Delete those columns first, and remove other people's names and contact details from club or family lists. You rarely need them to answer the question.

What never to tell an AI chatbot

How to stop AI chatbots using your chats for training

Ask specific questions

Start by asking it to describe the file, so you know it has read it properly. Then ask one clear question at a time.

Check it understands the file
Describe this file: what each column contains, how many rows there are, the date range, and anything that looks odd or missing. Don't analyse anything yet.
Where does the money go?
Group these transactions into sensible categories like groceries, eating out, transport, bills and subscriptions. Give me a monthly total for each category and a bar chart. List any transactions you weren't sure how to categorise.
Find subscriptions
Find payments that repeat every month or year, with the amount and the date they go out. Total them up.

Check the numbers

OpenAI recommends reviewing the code, outputs and assumptions when Python is used. Pick a total you can calculate yourself and ask which rows the assistant included or excluded.

An uploaded file can still be too complex to analyse completely. Ask how many rows and sheets were read and investigate a missing section instead of assuming a polished chart proves completeness.

Show the calculation
Show how you calculated [FIGURE]. State the sheet, rows and columns used, any exclusions and whether you ran code. Tell me if any of the file was not analysed.

Start with a small, anonymised file

Use made-up spending or remove personal details before uploading. Keep the original file so you can compare results.

  1. Prepare a CSV or Excel file with clear headers and one row per item.
  2. Use the attachment or file-upload control in ChatGPT to add it, if available to your account.
  3. Ask for the sheet names, row count and meaning of each column before requesting totals.
  4. Ask one calculation at a time and request the method or code used.
  5. Check a known total yourself, then inspect any differences before using the result.

Before you upload a spreadsheet

  • Delete account numbers, card numbers, sort codes and routing numbers
  • Remove other people's names and contact details
  • Keep only the columns you need
  • Know one total yourself, so you can check its answer

Sources

These sources support the guide. Tool settings, prices and policies change, so dated source boxes show when those details were checked.

AI helps with research and drafting. Dan checks the sources and is responsible for what is published.

How we check