OpenAI shipped ChatGPT Personal Finance for U.S. Pro users on May 15, and the pitch is exactly what you’d expect: connect your bank through Plaid, get a dashboard of spending and net worth, and then have a conversation with the same assistant you already use for code review and trip planning about whether you can actually afford the apartment. Twelve thousand-plus institutions are supported on day one — Schwab, Fidelity, Chase, AmEx, Robinhood, Capital One — and OpenAI says Intuit is the next big integration on the roadmap.
I’ve spent the last two weeks running it side by side with Copilot Money, Monarch, Rocket Money, and Cleo. The short version: ChatGPT Personal Finance is genuinely useful in a way none of the dedicated apps are, and it’s also worse than every single one of them at the things dedicated apps were built to do. Whether that trade is worth the Pro subscription depends almost entirely on what you actually want a money app for.
What OpenAI actually shipped on May 15
The product is a sidebar in ChatGPT. You authorize Plaid, pick your institutions, and within a few minutes you get a tiled dashboard of cash, credit, investments, and a “upcoming payments” strip pulled from recurring patterns. Categorization is handled by an LLM rather than a rules engine, which makes the first-pass labels surprisingly accurate on messy merchant strings — the kind of “SQ *AMARO COFF 8472” that Mint always called “Other” and Rocket Money still calls “Restaurants” for some reason.
The dashboard isn’t the headline, though. The headline is that you can now ask ChatGPT things like “if I take the Tokyo offer at ¥18M and my rent goes to ¥220K, what does my monthly cash flow actually look like after I keep my IRA contributions and the AmEx Platinum?” and it’ll pull from your real numbers instead of asking you to make them up. That single capability is the entire reason this product exists, and it’s the only reason any of the dedicated apps should be worried.
A few important boundaries. Connections are read-only — no transfers, no bill pay, no ability for ChatGPT to do anything to your accounts. It’s Pro-only for now ($200/month, not the $20 Plus tier), with Plus rollout described as “after feedback.” OpenAI’s data page says transactions are stored per your standard ChatGPT data controls and that model training opt-out applies; the product page is more cautious than the Twitter announcement was. There’s no mobile-first experience yet — the dashboard renders inside the web client and the iOS/Android apps, but it’s clearly built for desktop.
The five apps I’d actually compare it against
Skipping the dead — Mint is gone, Personal Capital became Empower and the free product withered, and the long tail of “AI budgeting” wrappers don’t have institutional connections that stay alive past quarter two. The contenders worth your time:
Copilot Money is the iOS-first one. Beautiful interface, $13/month or $95/year, no ads, no data sale, and the most enjoyable transaction-review experience in the category. The AI categorization is the original modern implementation — they shipped it before “AI” became a marketing checkbox.
Monarch Money is what the Mint refugees migrated to. Web and mobile, $15/month or $100/year, joint households as a first-class concept, and the best investment-tracking story outside the brokerages themselves. It’s the only one of these that genuinely handles couples-with-separate-accounts without making you compromise.
Rocket Money is the subscription killer. It’s free with a $4-12/month premium tier, but the real product is the team of negotiators that will cancel your gym membership for you and call Comcast for a discount. The budgeting is fine. The bill-negotiation is the moat.
Cleo is the conversational one — a chatbot that roasts you for buying DoorDash three nights in a row. Free with a $6-15/month plus tier, aimed squarely at sub-$50K earners trying to build a savings habit. The personality is real and the demographic fit is unusually tight.
Quicken Simplifi is the boring competent one I’d recommend to my parents. $48/year, no nonsense, calendar-based cash-flow forecasting that’s still better than anything the newer apps have copied.
Where ChatGPT Personal Finance actually wins
Free-form planning conversations are the unambiguous win, and they’re not close. I asked it to model a scenario where I took a six-month sabbatical, drew down emergency savings at $4,200/month, kept paying my Roth contributions, and projected what my net worth looked like at month seven assuming a 4% return on the brokerage. Copilot can’t do that. Monarch can’t do that. Rocket Money will literally tell you “I can help you cancel subscriptions” and nothing else. None of the dedicated apps will let you reason about your money in the language you actually think about your money in.
The categorization is also better than rules-based engines on the genuinely ambiguous stuff — irregular freelance income, weird Stripe payouts with reference codes attached, foreign currency charges that come through as the converted USD amount with a fee tagged onto a separate line. Copilot’s AI categorization is in the same ballpark and was there first, but ChatGPT will explain why it categorized something a particular way if you ask, and that turns out to matter when you’re reconciling.
The other quiet win is having one less app. If you already pay for ChatGPT Pro for work, the marginal cost of the finance feature is zero. Every other app on this list is a separate $10-15/month subscription, a separate login, a separate notification stream, and a separate place your transaction data lives.
Where the incumbents still beat it badly
Categorization rules and split transactions are not solved. Copilot lets you write “always categorize Amazon under household supplies unless the description contains ‘kindle’” and have it stick across thousands of future transactions. ChatGPT will do it once and then quietly forget by Tuesday, because there’s no persistent rule engine — it’s reasoning over conversation history. For people who care about precise category totals at the end of the year, this is a deal-breaker.
Investment tracking is still Monarch’s category. ChatGPT pulls balances and holdings, but the day-over-day performance views, the cost-basis tracking across multiple brokerages, the “what’s my actual asset allocation across the 401(k), the Roth, and the taxable account” rollup — Monarch nailed this years ago and the new entrant is at maybe 40% of that depth. If your net worth lives mostly in markets, this gap is huge.
Subscription cancellation is the Rocket Money moat and nothing else comes close. ChatGPT will identify your recurring charges and tell you what they cost; it will not call SiriusXM and argue with the retention agent for forty minutes. Rocket says it negotiates over $100M in bills annually and the average successful negotiation saves users around $720/year. That’s a real number. The AI assistant can flag the gym membership; the humans in Salt Lake City are the ones who cancel it.
Mobile is rough. The ChatGPT iOS app technically shows the dashboard but the UX clearly wasn’t designed for thumb-driven transaction review. Copilot is the gold standard for iOS-first budgeting and it’s not particularly close.
The privacy question I keep getting
Most of the questions I’ve gotten about this product are some variant of “is it safe to give ChatGPT my bank account.” The honest answer is “it’s about as safe as Plaid in general, which is the same trust boundary Copilot and Monarch and Rocket Money all sit behind.”
Plaid is a read-only data aggregator. None of these apps see your password — they see a token that lets them pull transaction history and balances. None of them, including ChatGPT, can move money. That’s worth saying out loud because the visceral reaction to “ChatGPT has my bank account” is doing a lot of work that the actual technical scope doesn’t justify.
The real difference is the data handling policy at the company that holds the Plaid token, and here OpenAI is the new entrant with the least track record. Copilot Money’s stance is the cleanest in the category — no ads, no data sale, paid product, transactions stay on their servers and don’t go anywhere. Monarch has the same shape. Rocket Money does monetize data — that’s part of why the base tier is free — and you should read their disclosures if that bothers you. Cleo is somewhere in between, with a paid tier that reduces the monetization footprint.
OpenAI says ChatGPT Personal Finance data follows your account’s data controls, that it’s not used for model training if you’ve opted out (and you can), and that Pro-tier conversations don’t go into the training pool by default. I believe them. I also note that this is OpenAI’s first consumer financial product, that the legal posture around financial data is materially different from chat history, and that the cleanest argument for not connecting it is “I don’t need a frontier model lab to be the seventh place my transaction data lives.”
If you’re already comfortable with one of the dedicated apps, the marginal trust cost of adding OpenAI is real. If you’re starting fresh and your alternative is “no app at all because I don’t want to deal with another subscription,” that calculus flips.
The accuracy reality nobody is warning new users about
LLM-driven categorization is better than rules engines on average and worse than rules engines on the tail. The average win shows up in fewer “Other” categories and smarter handling of weird merchant strings. The tail loss shows up when ChatGPT confidently tells you that you spent $847 on dining last month and you didn’t — it double-counted a Venmo reimbursement, or pulled in a pending charge that already got reversed, or treated a credit card payment as both a transfer and an expense.
I had it hallucinate a charge once during the two weeks. Not a categorization error — an actual transaction that wasn’t in any of my accounts, surfaced in a “where your money went” summary, with a confident dollar amount attached. When I asked it to show me the source transaction, it apologized and removed it. That’s a real category of failure that rules-based apps don’t have, because rules-based apps can only show you transactions that exist in the database.
The practical rule: don’t trust any single number an AI tells you about your money without being able to click through to the underlying transaction. ChatGPT’s dashboard does support drill-down on the tiles, which is good. The conversational summaries are where the hallucination risk lives, and the right discipline is to ask “show me the transactions behind that” before you act on any number.
Pick by what you actually want
If what you want is conversational planning — “can I afford this,” “what if I changed jobs,” “what does FIRE look like for me at 38” — ChatGPT Personal Finance is the first product that actually delivers on that promise. None of the dedicated apps can hold the conversation. Bolt this onto an existing rules-based budgeting setup and you have something genuinely new.
If what you want is the most enjoyable budgeting experience on iPhone, it’s Copilot Money. Still. The combination of haptics, transaction review flow, category rules that stick, and the original AI categorization implementation is two years ahead of everyone else on iOS specifically.
If what you want is shared household finance with real investment tracking, it’s Monarch. Couples with separate accounts, a joint budget, and brokerage rollups that survive a Schwab/TD merger is exactly the wheelhouse, and nobody else is close.
If what you want is someone to cancel your subscriptions and negotiate your bills, it’s Rocket Money and it’s not a comparison. The AI features are fine. The humans are the product.
If what you want is a coaching nudge to save $50 this week and you’re under 30, Cleo’s personality-driven UX still converts behavior change better than the spreadsheet-shaped competition. The category dismisses it because it doesn’t look serious; that’s exactly why it works for the audience it works for.
And if your alternative to all of this is a Google Sheet — that’s still a perfectly defensible answer. Most people who are good with money use one or two of these apps lightly and a spreadsheet seriously.
One thing to try this week
If you already pay for ChatGPT Pro and you’ve been on the fence: connect a single account — not all of them — and spend a week asking it the planning questions you’ve never been able to ask a budgeting app. Not “what did I spend on coffee,” which the dedicated apps answer better. Something like “if I want to save an extra $800/month for the next 14 months, where would that realistically come from in my actual spending?” That’s the conversation this product was built for, and a week with one account connected is enough to know whether it’s worth the full setup.
If you’re not already on Pro, don’t subscribe just for this. The Plus tier rollout is coming, and one of the other four apps probably fits your actual use case better at a fifth of the price.