Analyze your credit card spending with AI

What you need

• A computer
• A ChatGPT or Claude account (free plans limit file size and the number of uses)
• 1 to 3 transaction exports (CSV or Excel) downloaded from your bank or card website or app
• A spreadsheet program such as Excel (to check the results)

What you get

A summary and charts with spending share by category, the top 10 biggest charges, a list of likely subscriptions, and the change from last month. You can also download it as a cleaned-up spreadsheet. The example below uses fictional data.

Step by step

1. 1

   Download a transaction export from your bank or card

   Sign in to your bank or card issuer's website or app. Find the transactions or statement section and set the date range to this month. If there is a button to download a CSV or Excel file, use it and save the file. Menu names and locations differ by bank. If you cannot find it, search the bank's help center for "export transactions" or "download statement".

   You know it worked whenIf you open the downloaded file and see dates, merchant names, and amounts, it works.

   TipIf you have several cards, download each one separately so you have several files. Some mobile apps only offer PDF. In that case, use the bank's website on a computer.

2. 2

   Delete card number, account number, and name columns

   Open the file in a spreadsheet. If it has columns such as "Card number", "Account number", "Authorization code", or "Name", right-click the column header and delete them. For this analysis you only need date, merchant, and amount. Use Save As so the original stays untouched.

   You know it worked whenYou are ready when only columns like date, merchant, and amount remain.

   TipA card statement shows where you bought what. If some merchants are sensitive (such as clinics or pharmacies), you can delete those rows before uploading without hurting the analysis much.

3. 3

   Upload the file to the AI and ask for the analysis

   Sign in at chatgpt.com or claude.ai and upload the spreadsheet with the paperclip (attach) button in the chat box. Paste in the prompt below and send it.

   Prompt to pasteCopy prompt

   Analyze the attached credit card statement spreadsheet. It covers one month.
   
   What I want to know:
   1. Spending by category (Food, Transport, Shopping, Subscriptions, Health, Other): amount and percentage, as a table and a pie chart
   2. The top 10 biggest charges
   3. A list of charges that look like they repeat every month (possible subscriptions)
   4. Spending by weekday and by date as a bar chart
   5. Three places I could cut back (show the charges that support each one)
   
   If a merchant is hard to categorize, don't put it in "Other". Show those in a separate list and I will tell you the category.
   Don't include card numbers or personal information in the results.
   Also give me the cleaned-up results as a downloadable spreadsheet file.

   Make it yours

   • Change the categories to fit your life, for example Pets, Kids, or Hobbies.
   • To compare several months, add: "The 3 attached files are for August, September, and October. Also show the change month to month."
   • If item 5 (places to cut back) feels like too much, delete it.

   Want more control? (a slightly more technical version)

   Copy this version

   Write a Python script that merges and analyzes the attached card statement spreadsheets.
   - Read them with pandas. Even if column names differ between files, standardize them to date, merchant, and amount
   - Auto-categorize by keywords in the merchant name (keep the category table in a separate file so I can edit it)
   - Save the results as a monthly summary spreadsheet and PNG charts
   - Also tell me how to run it so that each week it only processes new files in the same folder

   You know it worked whenAfter a moment, the tables, charts, and summary appear together. If the charts show up as images, it works.

   TipIf no chart appears and it only describes the results in words, say "Make the charts as images." Free plans sometimes cannot make charts. In that case, take just the tables and build a chart in your spreadsheet.

4. 4

   Tell it how to categorize the unclear merchants

   When the AI shows a list of merchants it could not categorize, answer them one by one. For example: "Example Mart is Food, Corner Coffee is Food, anything I do not know is Other." Ask it to recalculate with the rules you gave.

   You know it worked whenIf the share of Other drops sharply and the category totals match the overall total, it works.

5. 5

   Change it by asking

   If you do not like the result, ask for changes in the same chat.

   TipTry things like "Split Food into Dining out and Groceries", "Use a horizontal bar chart instead of a pie chart", "Add the total savings if I cancel each subscription candidate", "Collect only charges over $100 in a separate list", or "Make the chart colors calmer".

6. 6

   Download the results spreadsheet and save it

   Say "Give me the cleaned-up results as a spreadsheet file" and a download link appears. Click it to save the file. You can save a chart as an image by right-clicking it.

   You know it worked whenIf the spreadsheet has the category table and the top 10 on separate sheets, it works.

   TipDouble-check the total once. In a spreadsheet, select the amount column and the sum shows in the status bar at the bottom. It should match the total the AI reported.

7. 7

   Continue next month

   When next month's file is ready, upload it in the same chat and say "Tell me what changed from last month." If you started a new chat, upload both months together.

   You know it worked whenIf the change by category appears as a table, it works.

If you get stuck

Try next

• If you want to repeat this every month, get an analysis script with the advanced prompt. It runs on your computer, so you may need a developer's help.
• To save your category rules, put a notes file with the rules and your past statements into the Projects feature of ChatGPT or Claude. Then you do not have to explain everything from scratch next month.

Features used here

More recipes