What do AI assistants tell people about local banks and credit unions?

What did we ask, and where?

Among Americans looking for financial guidance, about one in four Gen Z and millennials have asked an AI tool (Gallup, 2026). So we asked ChatGPT, Gemini and Claude the questions people ask about money in Denver: where to open a savings account, which credit unions they can join, the best certificate rates, free checking, auto loans, a retiree with certificates maturing. Hundreds of answers, gathered on October 4, 2026.

Denver has more than 70 banks and credit unions with at least $500 million in assets and a branch in the metro. We counted which ones the answers named.

Who gets named?

A few names take most of the answers. The most-named credit union appeared in seven in ten answers. Credit unions with Denver branches appeared in nine in ten; large national banks with Denver branches in three in ten.

About a third of Denver's banks and credit unions over $500 million were never named at all, in any answer, from any assistant. For them, a growing share of first questions about money is a conversation they are not part of.

How current are the answers?

Less current than they sound. About one in five answers recommended an institution under a name that no longer exists, without saying so: a bank that merged into a national bank in June 2026, a credit union that merged in March 2025, and names retired years ago. In other answers, the same assistants did note some of those changes, so the right information was available. It was not used every time.

Rates were quoted often and dated rarely. Seven in ten answers quoted at least one rate, and six in ten of those did not say when the rate was true.

Do people get the same answer twice?

Often not. Asked the same question again, the same assistant named different institutions: only about three in ten were named every time. The three assistants agreed on the most-named local institution for about one in nine questions. An institution named once may not be named the next time someone asks.

Where do the answers send people?

Comparison and personal-finance sites carry a large share of what the assistants read. And some citations led to a credit union in another state whose name is one letter away from a Denver credit union's. Someone clicking through to check membership would land on the wrong institution.

What can an institution do?

Assistants repeat what they can read. The steps below make an institution easier to read correctly. None of them guarantees a place in an answer.

• Say who you are, plainly, on your own site. Your current name, any former names and mergers, where you have branches, and who can join. An assistant that finds an old name on a comparison site needs your page to correct it.
• Date your rates. A rate page with a visible date is easier to trust and harder to quote out of date.
• Make eligibility a page, not a PDF. Questions about who can join are among the most common, and the answer should be readable text.
• Check the comparison sites. Make sure the listings that assistants read carry your current name and products.
• Follow the link. Where an answer sends someone, that page should let them act: open the account, apply, or book a time.
• Measure it over time. Answers change from one asking to the next, so a single test tells you little.