ChatGPT can now track your money

Most people manage their money through a messy patchwork of banking apps, credit card portals, and forgotten spreadsheets, rarely getting a clear picture of what they can actually spend next month.

OpenAI just launched an early preview of personal finance tools directly inside ChatGPT for U.S. Pro subscribers, rolling out on web and iOS.

The headline feature: you can now link your real-world bank and investment accounts directly to the chatbot.

12,000 banks in a single chat box

OpenAI partnered with financial infrastructure provider Plaid to support connections to over 12,000 financial institutions, including Chase, Fidelity, Charles Schwab, Robinhood, American Express, and Capital One. Support for Intuit is also in the pipeline.

Users can get started by clicking "Finances" in the sidebar or simply typing "@Finances, connect my accounts" into any chat. After authenticating through Plaid, ChatGPT syncs and categorizes transactions in a few minutes.

Once connected, users get a clean dashboard covering portfolio performance, spending habits, recurring subscriptions, and upcoming bills.

More importantly, you can cross-examine your own balance sheet using natural language. Instead of manually calculating figures, you can ask prompts like "I feel like I've been spending more recently, what changed?" or "Help me build a plan to buy a home in five years."

Under the hood, financial conversations default to the GPT-5.5 Thinking reasoning model. OpenAI worked with more than 50 finance professionals from top institutions to build an internal benchmark specifically for grading answers on complex personal finance tasks.

The launch comes just one month after OpenAI acquired the team behind Hiro, a personal finance startup backed by Ribbit and General Catalyst, whose domain expertise clearly helped shape the release.

Dedicated financial memories, but no transaction access

ChatGPT is also rolling out "Financial memories," a dedicated memory tier tailored for long-term planning.

Beyond raw bank transactions, users can volunteer life context that does not show up on statements, like "I plan to buy a car early next year" or "I owe my parents money." The model retains these details to inform future chats without needing a refresher.

Handling sensitive financial data naturally raises red flags, and OpenAI set clear operational boundaries. ChatGPT can see balances, transactions, investments, and liabilities, but it cannot view full account numbers or initiate transfers.

On the privacy side, temporary chats block access to financial accounts. If you disconnect an account, OpenAI purges synced records within 30 days, and financial memories can be reviewed or deleted anytime.

OpenAI noted that over 200 million people already ask ChatGPT financial questions every month. The company later expanded the rollout to Plus users in late June, brought it to Free and Go users in October, and gradually added credit score tracking, stock watchlists, voice controls, and manual account logging.

Why it matters

General-purpose chatbots are aggressively pushing into deep, high-stakes domains like health and wealth management.

For regular users, the real win here is context-aware reasoning. Traditional budgeting apps show you that you spent $400 on dining out, but they cannot simulate what happens to your runway if you quit your job in three months based on your actual debt and savings. ChatGPT can. That said, handing over raw bank feeds to an AI platform requires massive trust, and privacy-conscious users will likely hesitate to grant full visibility via Plaid.

From an industry perspective, partnering with Plaid and Intuit is a brilliant shortcut. Instead of navigating brutal banking compliance rules alone, OpenAI leverages Plaid for plumbing and Intuit for high-intent actions like filing taxes or checking credit card approval odds. This poses an existential threat to standalone budgeting and net-worth tracking apps. When the AI you already use every day can ingest your actual bank data and answer financial questions on the fly, niche utilities lose their purpose.

However, we think there is an obvious lingering risk. OpenAI clearly states this is not professional financial advice, but when a reasoning model crunches real personal balance sheets and suggests actionable timelines, users will inevitably treat it like one. If the model miscalculates tax implications or assumes unrealistic investment returns, real-world mistakes will hurt much more than a generic hallucination.