You’ve probably already done it.

Late at night, staring at your bank balance, you opened ChatGPT, Claude or Gemini and typed something like, “How should I save for retirement?” And in about three seconds, you got a polished, confident answer that sounded like it came from a financial planner charging $300 an hour.

Here’s the problem: That answer might’ve been dead wrong.

An Intuit Credit Karma survey found that two-thirds of Americans who’ve used generative AI have asked it for financial advice. Among Gen Z and millennials, that number hits 82%. And 85% of them actually followed the advice they received.

But according to a Fortune analysis of the same data, over half of respondents who followed AI money advice made a poor financial decision because of it.

So millions of people are using a tool that’s helpful roughly half the time and harmful the other half. That’s not a financial strategy. That’s a coin flip.

Or maybe they’re just not doing it right.

I use AI all the time, including to explore investment ideas.

Example: Last November, I wrote an article called “Anatomy of an AI Trade: How I Use AI to Make Money in Stocks.” In that article I wrote about using AI to help me find an exchange-traded fund (ETF) that offers exposure to small- and mid-cap biotech stocks. It suggested the ALPS Medical Breakthroughs ETF. After doing a bit more research, I bought some.

As I write this, that ETF is up more than 23%.

I believe AI can help investors, providing they’re using it correctly. And anyone can learn how: It just takes practice.

Here are nine tips that are going to help you harness the power of AI to manage your finances.

1. Stop asking vague questions

This is the single biggest mistake people make, and it’s where an MIT professor’s research gets really interesting.

Andrew Lo, director of MIT’s Laboratory for Financial Engineering, told CNBC that most people type something lazy like “How should I retire?” and expect useful output. That’s like walking into a doctor’s office and saying, “Fix me,” without describing your symptoms.

Lo’s blunt assessment? “Garbage in, garbage out.”

The quality of your answer depends almost entirely on the quality of your question. And most people’s questions are terrible.

2. Use the fiduciary hack

Here’s Lo’s fix, and it’s brilliant.

Instead of asking a vague question, tell the AI to assume a specific role. Lo’s example prompt goes something like this: “Assume you are a fee-only fiduciary financial advisor. Here are my goals, constraints, tax bracket, state, assets, risk tolerance, and timeline.”

Then ask for a base-case strategy, key assumptions, risks, what could invalidate the plan, and — this is the critical part — what information the AI is missing or uncertain about.

By forcing the chatbot to act as a fiduciary, you’re essentially telling it to prioritize your interests and flag its own blind spots. It doesn’t make the AI legally liable, but it dramatically changes the quality of what comes out.

3. Know where AI is genuinely useful

In my experience, AI chatbots are good at explaining financial concepts in plain English. Things like how a Roth IRA conversion works, why diversification matters, or the difference between a tax credit and a deduction.

They’re also great for drafting scripts to negotiate your bills, analyzing your spending if you paste in a bank statement, or comparing the features of two financial products side by side.

Think of it this way: AI is a spectacular research assistant. It’s a lousy decision-maker.

4. Never trust it with your taxes

This one could actually cost you money — real money, with Internal Revenue Service (IRS) penalties attached.

About 26% of Americans used AI to help file their 2025 tax returns, more than double the prior year, according to an Adobe poll. Some users reported getting bigger refunds. Others got bad advice.

Here’s what most people don’t realize: The IRS doesn’t care that a chatbot told you to claim a deduction. If the AI is wrong, that’s your mistake. There’s no “the robot ate my homework” defense.

Tax law changes constantly. The latest changes under the new tax bill may not even be reflected in AI responses yet. As a CPA myself, I can tell you that taxes are the last place to trust unverified AI output, at least for now.

I have no doubt, however, that the day will soon come when AI will be able to do your taxes. If I were working in the tax preparation business, I’d be thinking of my next career.

5. Understand the 35% problem

A study by broker analysis site Investing in the Web asked ChatGPT 100 finance questions and had the answers reviewed by industry experts. The result? According to Entrepreneur’s reporting, 35% of the answers were wrong or misleading.

That’s not 35% obviously wrong. That’s the sneaky kind — answers that sound authoritative and polished but leave out critical details or cite outdated information.

As a Purdue University accounting professor told CNBC, AI will confidently convince you that the sky is green. The mechanics of the answer look perfect, but the conclusion is flat-out wrong.

Important to note: That story about the 35% error rate published a year ago, which is like a century in AI time. Nonetheless, be aware that the confidence expressed by AI may be misplaced. Verify everything.

6. Ask it what it doesn’t know

This is the single most underrated trick in AI prompting, and it comes straight from Lo’s playbook.

After you get your answer, follow up with: “What information are you uncertain about? What data is missing? What could make this advice wrong?”

A good financial advisor tells you the risks. A great one tells you what they don’t know. Most humans won’t volunteer that information, but AI actually will — if you ask.

This step separates the people who get value from AI financial tools from the people who get burned.

7. Don’t hand over your financial life

Three-quarters of people who use AI for financial advice say it lets them ask questions they’d be too embarrassed to ask anyone else, according to that Intuit Credit Karma survey. That’s actually a benefit — no judgment, no sales pitch, available at 2 a.m.

But the flip side is that people are sharing incredibly sensitive data with chatbots, such as income, debts, Social Security numbers, and full tax documents.

Ask yourself: Do you know where that data goes? Who trains on it? Whether it’s stored or deleted?

Use AI for general questions and concept explanations. When it comes to sharing your complete financial picture, be aware that all that information is potentially being absorbed by the company behind the AI.

8. Use the reverse-prompt shortcut

Here’s another gem from Lo that almost nobody knows about.

After you’ve gone back and forth with the AI through 10 or 20 prompts and finally gotten a really solid answer, ask it one more question: “What prompt should I have asked you in the first place to get this answer?”

The AI will spit out an optimized prompt you can save and reuse next time. It’s like getting a custom key made for a lock you’ve already picked the hard way.

Over time, you build a small library of prompts that skip the garbage and go straight to the good stuff. That’s how the people getting real value from AI are doing it.

9. Watch what’s coming next

The big tech companies aren’t just dabbling in financial AI. They’re going all in.

OpenAI just acquired a personal finance startup called Hiro Finance — the company’s second fintech acquisition in six months, according to TechCrunch. Hiro’s vision was an “AI personal CFO” that could manage household budgets, optimize tax strategies, and execute financial plans.

As American Banker reported, one fintech analyst called the move an aggressive push into financial services, aimed at capturing “share of mind” in how consumers think about money.

As I mentioned above, the tools are going to get dramatically better, fast. But “better” doesn’t mean “perfect,” and it definitely doesn’t mean “accountable.” An AI can’t be sued for bad advice. It can’t be fined by the SEC. It doesn’t lose sleep if it wipes out your savings. So use it, but do so carefully.

The bottom line

AI is the most powerful free financial tool ever created. It’s also the most dangerous one because it sounds like an expert even when it’s wrong.

Use it to learn. Use it to research. Use it to draft, compare, and brainstorm.

But when it’s time to make a real decision with real money? Verify everything. Cross-check the numbers. And for complex situations involving taxes, estates, or retirement income, get a qualified human advisor who has a legal obligation to put your interests first.

AI doesn’t have a fiduciary duty. It doesn’t have skin in the game. And it definitely won’t cover your losses when it gets something wrong.

You will.

Source link