ChatGPT's Finance Preview Depends On Account-Level Trust
OpenAI's personal finance preview connects ChatGPT to bank and investment data for US Pro users, raising the value and the risk of AI advice at the same time.
OpenAI has started previewing a personal finance experience in ChatGPT for Pro users in the United States. The feature lets users connect bank and investment accounts, view a money dashboard and ask questions that use their actual balances, transactions, liabilities, goals and financial memories.
The rollout is limited to web and iOS at launch. OpenAI says it supports more than 12,000 financial institutions through Plaid, with Intuit support planned, and that the feature will expand beyond Pro after the company learns from early use. GPT-5.5 Thinking is the default model for connected finance conversations, with GPT-5.5 Pro available for users who want the highest-reasoning option.
The product direction is clear: OpenAI wants financial help in ChatGPT to move from generic advice to contextual analysis. A user asking how to save more money currently has to describe income, spending, debts and goals well enough for the model to respond usefully. Connected accounts remove some of that manual explanation and let the assistant inspect the underlying patterns directly.
OpenAI’s examples show questions about spending trends, subscription costs, savings goals and investment planning. That kind of assistance can be valuable when the model is working from real transaction categories rather than a rough prompt. It also moves ChatGPT closer to the territory occupied by budgeting apps, personal finance dashboards and advisory tools, where trust and regulatory boundaries matter more than conversational fluency.
The company is trying to address those boundaries explicitly. It says ChatGPT cannot see full account numbers and cannot make changes to connected accounts. Users can disconnect accounts, after which synced account data is deleted from OpenAI’s systems within 30 days. Temporary chats do not access connected financial accounts, and financial memories can be viewed or deleted from the Finances page.
Those controls are necessary because the product becomes more useful as it receives more sensitive context. Financial data describes income, family obligations, debt pressure, risk appetite and life plans. If users come to rely on ChatGPT for budgeting, mortgage decisions or investment planning, OpenAI will need to keep the distinction between educational guidance and regulated advice clear.
OpenAI says it worked with more than 50 finance professionals to evaluate complex personal finance responses. On its internal benchmark, GPT-5.5 Thinking scored 79 out of 100 and GPT-5.5 Pro scored 82.5. Those figures provide some evidence of response quality, although they cannot fully predict how people will interpret advice when they are under financial stress or dealing with incomplete information.
For builders, the preview is another example of consumer AI products becoming valuable through integrations rather than conversation alone. The hard product work is permissioning, memory, reversibility, data retention, source transparency and user confidence. In finance, those details are not secondary settings. They are the product.
OpenAI’s finance preview will be judged on whether it can combine useful account-level analysis with controls that users and regulators trust. A clever answer in an empty chat window is easy to demo. A financial assistant connected to real accounts has to earn confidence every time it explains where a recommendation came from.
Published: 2026-05-16 - Sources: OpenAI