21 ChatGPT Prompts for Budgeting: A Monthly Money Review
21 copy-paste AI prompts for a monthly money review in ChatGPT, Claude or Gemini: spending drift, subscriptions, savings rate, and how to check the math.
Short answer: give the AI one month of your transactions, paste a setup prompt with your currency, household, pay cycle and goal, then ask one question at a time: where the money went, what changed, what repeats, and what to do next month. Ask it to show the sums, and check one total yourself. The 21 prompts below work in ChatGPT, Claude, Gemini or an on-device assistant.
Some people already use AI for money questions, but few trust it fully. A Gallup survey with Edward Jones asked 5,075 US adults about financial guidance. Of those who sought guidance in the past year, 18% used AI tools. Only 3% of US adults have a great deal of confidence in AI as a source of financial guidance (Gallup, August 2026). A monthly review fits that gap. The AI sorts and summarizes, and you check the numbers.
Key Takeaways
- Real data beats general tips. Share an export, connect your budget app with read-only access, or use an on-device assistant.
- Start every review with a setup prompt: context, a goal, tables, and permission to say “I don’t know”.
- Ask about one thing at a time: the month at a glance, category drift, recurring charges, savings rate, one-off costs, next month.
- AI can get sums wrong. Ask for the transactions behind a total and re-add one category yourself.
- Never share account numbers, card numbers, passwords or ID numbers.
Before you start: give the AI your data
A prompt without your numbers gets general tips back. There are three ways to share one month of transactions:
| Way | How it works | What leaves your device |
|---|---|---|
| Paste or upload an export | Export the month as a CSV file, remove identifying details, share it in the chat | The rows you choose to share |
| Connect your budget app | The AI looks up transactions through a read-only connection | The transactions the AI looks up |
| On-device assistant | The AI model runs on your phone | Nothing |
For the comparison prompts (4, 5, 7, 8, 11, 13 and 16), include the three earlier months as well.
An export gives you the most control. You decide which rows and columns the AI sees, so delete notes, names and anything you don’t need first. The data still goes to the AI company and is handled under its privacy terms. Some chatbot users already do this. PensionBee surveyed 1,000 US adults who already use AI chatbots for personal finance, and 32% had shared monthly spending or budget details with a chatbot (PensionBee, September 2026). The survey comes from a retirement provider and does not represent all Americans.
A connection saves the copy-paste. The AI can also reach earlier months, which helps with comparisons. Read-only means it cannot add, edit or delete anything. What it reads is still sent to the AI provider.
An on-device assistant keeps everything on the phone. The trade-off is a smaller model. It suits short, single questions more than a long multi-step review.
Whichever way you choose, never share account numbers, card numbers, passwords or ID numbers such as a Social Security number. A spending review does not need them.
The setup prompt
Paste this once at the start of each review, then add your data:
You are helping me review my household spending for [month and year].
Context:
- Currency: [USD]
- Household: [2 adults, 1 child]
- Pay cycle: [paid every two weeks]
- Main goal: [save $400 a month for an emergency fund]
- Date format in my data: [day/month/year]
Rules:
- Use only the data I give you. If something is missing or unclear,
say "I don't know" and tell me what you need.
- Show totals in tables.
- Show the sums behind every total so I can check them.
- Do not give investment, tax or legal advice.
- I will ask one question at a time. Answer only that question.
Reply "ready" and list any questions you have about the data.
Most parts follow published guidance from the AI companies:
| Part of the prompt | Why it is there |
|---|---|
| Currency, household, pay cycle, goal | OpenAI advises giving the model any extra information it might need to respond (OpenAI). Anthropic puts it this way: “The more precisely you explain what you want, the better the result.” (Anthropic) |
| “I don’t know” | Anthropic recommends allowing the model to say “I don’t know” to reduce made-up answers (Anthropic) |
| Tables | “Be specific about the output format and constraints you want.” (Anthropic). Google also notes you can ask for a table or a bulleted list (Google) |
| One question at a time | Google suggests breaking complex tasks into simpler parts (Google) |
| Show the sums | This guide’s own advice. See “Check the math” below |
21 prompts for your monthly money review
Replace the parts in [brackets]. Ask the prompts one at a time.
The month at a glance
A good answer is one small table whose totals match your budget app.
1. Month summary
Summarize [month]: total income, total spending and the difference. Then list my five largest categories with the amount and share of spending for each, in a table.
Example with made-up numbers. Prompt 1 on this data:
date,time,for,amount,description
01/10/2026,09:00,salary,4000,
03/10/2026,18:20,rent,-1500,
07/10/2026,12:10,groceries,-720,
12/10/2026,08:00,car,-480,repair
20/10/2026,20:15,eating out,-400,
28/10/2026,07:30,subscriptions,-300,
A good answer: income $4,000, spending $3,400, difference $600.
| Category | Amount | Share of spending |
|---|---|---|
| Rent | $1,500 | 44.1% |
| Groceries | $720 | 21.2% |
| Car | $480 | 14.1% |
| Eating out | $400 | 11.8% |
| Subscriptions | $300 | 8.8% |
2. Largest transactions
List my 10 largest transactions in [month] with date, name and amount. Mark any that look like duplicates.
3. Fixed vs everyday
Split [month]’s spending into fixed bills, everyday spending and everything else. Tell me which rows you were unsure about.
Category drift
A good answer names the transactions behind a change, not only a percentage.
4. Month over month
Compare each category in [month] with [last month]. Show both amounts and the change in [currency] and in percent, sorted by the largest increase.
5. Against the three-month average
Compare [month] with the average of the previous three months for each category. Flag categories more than [15]% above their average.
6. Explain the biggest jump
For the category that grew most, list the transactions that explain the increase. Was it more purchases, larger purchases, or one big item?
Subscriptions and recurring charges
A good answer is a list with dates and amounts that you can match to your records. Then decide what to keep with the guide to finding and canceling forgotten subscriptions.
7. Find what repeats
Find charges that repeat at about the same amount in [the last three months]. List the name, amount, how often it repeats and the yearly cost.
8. Price increases
Which recurring charges went up compared with earlier months? Show the old and new amount.
9. Small charges
List recurring charges under [$20] that are easy to overlook. Total them per month and per year.
Savings rate
The formula is savings rate = (income − spending) ÷ income. For example, with $4,000 of income and $3,400 of spending, the rate is $600 ÷ $4,000 = 15%. A good answer shows all three numbers. For a target, see how much to save each month.
10. This month’s rate
Calculate my savings rate for [month] as (income − spending) ÷ income. Show the income total, the spending total and the division.
11. The trend
Show my savings rate for each of the last [three] months in a table. Say which categories explain the best and the worst month.
12. Gap to my goal
My goal is to save [$400] a month. Based on [month], how far am I from it? Which two categories could close the gap? Show the numbers.
One-off and unusual expenses
A good answer separates one-time costs from regular spending, so a car repair does not distort the month. Costs that return every year belong in a sinking fund.
13. Spot the one-offs
List expenses in [month] that do not appear in earlier months and are larger than [$100]. Label each as likely one-time or likely to return.
14. The month without them
Recalculate [month]’s total spending without the one-time items. Show both totals side by side.
15. Yearly costs
Which of these costs come back every year, such as insurance, gifts or car repairs? Suggest a monthly amount to set aside for each, and show the division.
Plan next month
A good answer builds the budget from your real averages, and every number can be traced back.
16. Draft budget
Suggest a budget for [next month] by category, based on my three-month averages. Show the average next to each suggested amount and explain any difference.
17. Known changes
Known changes next month: [rent rises by $50; a birthday; an annual insurance bill]. Adjust the budget and show what changed.
18. Three specific changes
Give me three specific changes for next month. For each, show the category, the amount saved and the transactions it is based on. No general tips.
Couples and shared budgets
These work when your data shows who paid for each item. A good answer reports numbers and stays neutral. For the setup itself, see shared budgets for couples.
19. Who paid what
This data covers two people: [Name A] and [Name B]. Show spending by person and by category in a table. Report the numbers only, without judging either person.
20. Shared costs
List our shared costs (rent, utilities, groceries) and what each of us paid in [month]. Then show what a [50/50] split would have been, and the difference.
21. Money talk agenda
Write a short agenda for a 15-minute money talk: two numbers to discuss, one thing that improved and one decision to make for [next month]. Base each item on the data.
Check the math
AI chatbots can present a wrong number as fact. US regulators said so in a 2024 investor alert about AI and investing: “Even when based on accurate input, information resulting from AI can be faulty, or even completely made up.” (SEC, NASAA and FINRA). Google says the same about its own product: “LLM experiences (Gemini Apps included) can hallucinate and present inaccurate information as factual.” (Google).
Many chatbot users say they would skip the check. In the PensionBee survey of 1,000 US adults who use AI chatbots for personal finance (PensionBee, September 2026), 23% said a chatbot had given them incorrect financial information, and 57% would accept a chatbot’s answer on a financial decision without checking.
Three quick checks make errors easier to spot:
- Ask for the rows. “List every transaction you counted in [category], with amounts, and add them up step by step.”
- Re-add one category yourself. Use a calculator or your budget app’s category total. If one total is off, distrust the rest.
- Ask for assumptions. “List every assumption you made, and every row you skipped or could not read.”
Anthropic’s guidance for its own model is direct: “Always validate critical information, especially for high-stakes decisions.” (Anthropic).
A 20-minute monthly routine
Close tracking is not the norm. Asked how they manage their finances each month, 18.4% of 1,000 US adults said they set a detailed monthly budget and track most income and expenses, while 28.0% have a general idea but do not formally track (Verasight, May 2026). A fixed routine keeps the review short enough to repeat:
- Get the data ready (3 minutes). Export the month, or open the connected chat.
- Paste the setup prompt (2 minutes). Answer the AI’s questions about the data.
- Look at the month (5 minutes). Run prompts 1, 4 and 5.
- Find repeats and one-offs (4 minutes). Run prompts 7 and 13.
- Check the math (3 minutes). Re-add one category.
- Plan next month (3 minutes). Run prompts 10 and 16, and write down one change.
Add the other prompts when something looks off.
Doing this in Japka
Japka supports all three ways of sharing data:
- Export: Settings → Export CSV is free. The file has the columns
date,time,for,amount,descriptionandlocation, with no category column, so ask the AI to group rows by theforname. Dates are day/month/year, so tell the AI that. The CSV guide shows the format. - Connect: Japka Premium includes a read-only connection to ChatGPT (or your favorite AI chat). Japka adds up totals on its own server, so those totals do not depend on the AI’s arithmetic. The data the AI reads goes to that AI provider.
- On-device: the AI tab in Japka Premium runs an AI model on your phone, offline, with nothing uploaded.
The on-device model is smaller than cloud models. Use it for short, single questions, and use a connected cloud AI for long, multi-step prompts like 5, 16 and 20.
The bottom line
A monthly money review with AI works when it runs on your real transactions and you verify the result. Share only the data the review needs, start with the setup prompt, ask one question at a time, and re-add one total yourself. Twenty minutes a month is enough to see what changed and pick one thing to do next.
Written and maintained by the Japka team, the makers of the Japka budget app. Read how we write our guides, or contact us with corrections.
This guide is general information, not financial advice.
FAQ
Is it safe to share my budget with ChatGPT, Claude or Gemini?
It depends on what you share. Category totals and transaction rows without names are lower risk than full statements. The data is handled under the AI company's privacy terms, so read them first. Never share account numbers, card numbers, passwords or ID numbers.
Do these prompts work in Claude and Gemini too?
Yes. The prompts are plain English and do not depend on one chatbot. What matters more is how the AI gets your data and whether you check its totals.
Can AI chatbots make mistakes with my numbers?
Yes. They can miscount rows, skip lines or invent a figure. Ask for the transactions behind each total, and compare one category with your own records before you act on the answer.
Can an AI chatbot replace a financial advisor?
No. A chatbot can help you see where your money went. Google's own guidance says: "Don't rely on Gemini Apps for diagnosis, treatment, medical advice, or legal, financial, or other professional help." For investments, taxes or debt problems, talk to a qualified professional.