OpenAI Group PBC has added a set of personal finance features to ChatGPT Pro, the priciest consumer tier of its popular chatbot service.

On launch, the capabilities are available in preview to a limited number of U.S. users. OpenAI will collect feedback from the testers to refine the features before making them more broadly accessible. The company stated in a blog post today that it plans to make the features “available to everyone,” which hints they may eventually roll out to free accounts.

Before consumers can start using ChatGPT for personal finance tasks, they have to connect it to their bank accounts via Plaid. It’s a service operated by a startup of the same name that helps developers integrate their applications with financial institutions’ systems. OpenAI says that Plaid enables ChatGPT to pull data from more than 12,000 banks.

Once all the necessary integrations are set up, ChatGPT creates a dashboard that shows key information from the user’s bank accounts. Simple graphs visualize monthly spending numbers, the asset types that make up the user’s investment portfolio and other high-level details. ChatGPT Pro subscribers can dig deeper by asking natural language questions.

One of the tasks that the chatbot promises to ease is finding ways to improve the user’s finances. A consumer could ask it to identify subscriptions that can be safely canceled or check for risky stocks in an investment portfolio. Additionally, ChatGPT can perform tasks that require analyzing data from multiple accounts and credit cards. For example, it could add up the expenses incurred during a vacation to calculate the total price tag.

The chatbot is capable of analyzing not only individual financial events but also trends. It can highlight how a user’s monthly spending fluctuated over the 12 months and generate multiyear budget roadmaps. A consumer looking to buy a new car could ask ChatGPT to create a savings plan that makes it easier to realize that goal.

The new personal finance features are powered by GPT-5.5 Thinking, a version of OpenAI’s latest large language model that is optimized for reasoning. It scored 60% on a benchmark called FinanceAgent that measures LLMs’ ability to perform tasks such as analyzing earnings reports. GPT-5.5 set a record on a second evaluation, FrontierMath Tier 4, that comprises doctorate-level math challenges.

OpenAI created a custom personal finance benchmark with the help of more than 50 financial professionals to measure ChatGPT’s performance. The evaluation will help its engineers improve the chatbot’s performance over time. 

According to OpenAI, future versions of ChatGPT will be capable of taking a more active role in users’ personal finance decisions. Currently, a consumer planning to apply to a new credit card can only ask the chatbot for guidance. In the future, ChatGPT will be capable of submitting the card application and estimating its odds of receiving approval. OpenAI also sees consumers using the chatbot for a wide range of related tasks.

Image: OpenAI