ChatGPT Finances: what it does, where it stops, and what to use instead.

OpenAI launched Finances for U.S. Pro users in May 2026, expanded to Plus in June, and began rolling it out to Free and Go on October 2. OpenAI's help center now lists it for Free, Go, Plus, and Pro users in the U.S. on web, iOS, and Android.

On Plus and Pro, OpenAI began rolling out a separate Experian connection on September 21, 2026, which shows your credit report and VantageScore 3.0 credit score, updated monthly, though some eligible users may not see it yet. OpenAI's help center says checking it is a soft inquiry that does not affect your score, and you can disconnect it without removing your bank accounts.

OpenAI says the connection cannot change your accounts. Financial memories can preserve context you share. The current announcement is linked below.

Ask a focused question, then check the records and assumptions behind the response. OpenAI says the feature does not replace professional financial advice. Its model benchmark measures response quality and accuracy together, not a guarantee about any individual calculation.

Weekly updates mean it is no longer accurate to describe ChatGPT Finances as only answering when asked. The published model benchmark also does not establish how every dashboard total is calculated.

Each item gives your agent a place to start. A possible duplicate charge comes with evidence; your agent checks whether it is a problem, discusses it with you, and can save a follow-up. Candor picks the starting points. Your agent decides what to recommend and acts through its own authorized tools.

The practical difference is who starts the conversation. Every week Candor reviews your records and notes and raises what's worth talking through, from a subscription whose price just rose to cash sitting idle, budget drift, or an employer match left on the table, each with the records behind it. When nothing matters, it raises nothing. Your agent reads the review when you ask, or on a schedule if it supports scheduled tasks, say every Monday morning, and brings you the decision. Nobody has to remember to ask.

The math difference is architectural. Candor computes totals, budget pace, and variance server-side from the transactions and returns them with the records behind them, so the agent reports a figure instead of predicting one. The trust difference is the same shape: Candor holds no money-movement tools at all, is read-only toward your institutions by design, logs every read with a reason, never sells or trains on your data, and lets you export your data and delete it. Your agent still makes the judgment calls and acts under your authority; it just does so from a workspace that remembers.

Link your accounts in Candor rather than in both places. One bank connection is one less to remove later, and ChatGPT reads what it needs through Candor Finance. The setup page below covers the routes, none of which needs a terminal.

For example, ask your agent to review recurring charges in Candor, save which ones you intend to keep, and revisit a promised refund next time. Another connected agent can pick up that saved work. Scheduled follow-up depends on the agent you use and the schedule you set.

One documented case shows Grok Bot investigating overlapping subscriptions and securing a $1,149.69 refund after contacting support outside Candor. That is one outcome, not an expected return.

The comparison distinguishes published features from our assessment of fit. Rollouts can vary by account, so check the linked source and the feature in your own account.