Somewhere in your field of membership, a 34-year-old with a 13-year-old Honda opened an app last night and typed something like this:
"My car is falling apart and I have decent credit. Where should I get an auto loan?"
She got an answer. A good one, probably. Confident, organized, with three or four names in it and a short reason for each.
Your credit union was almost certainly not one of them.
Nobody searched. Nobody clicked. Nobody saw your billboard, your rate table, or the Facebook post your marketing coordinator spent two hours on. A model that has never set foot in your branch made a recommendation about who should hold that member's loan, and you were not in the room.
That is the whole thing. That is the article. Everything below is me proving it.
Here are the numbers, because I think we have all had enough of vibes-based strategy.
Zero-click searches hit roughly 68% of U.S. queries in early 2026, up from about 60% two years earlier, according to SparkToro. When Google shows an AI Overview, click-through to the traditional results drops by close to 60%. Seer Interactive tracked millions of queries and found organic click-through on AI Overview queries fell from about 1.76% to 0.61%.
Now the part that is our industry specifically.
A J.D. Power survey of 4,000 consumers found 51% of Americans use AI to get financial advice or information. TD Bank's annual survey found 55% had asked a large language model for financial advice, up from 10% the year before. Among Gen Z in that study, it was 77%. EY surveyed 18,000 people across 23 countries and found roughly half had used AI for a financial decision in the past six months.
Forrester's 2026 prediction is the one I would put in front of your board: financial institutions should expect human visits to their websites to fall by about 20% while machine-initiated traffic climbs.
Fewer people are coming to your site. More machines are. And almost every credit union website in America was built entirely for the first group.
Here is where most of the industry is getting it wrong, including some very expensive consultants.
Ranking on Google was a contest between your page and other pages for one keyword. You could see the keyword. You could see your position. You could watch it move.
Getting named in an AI answer works differently. The model is not picking a page. It is deciding whether your institution exists as a trustworthy, well-defined thing in the world, then assembling a recommendation from whatever it can corroborate about you across dozens of sources at once. Your own website is one input, and often not the loudest one.
What actually moves the needle is a mess of things you do not fully control. Third-party roundups and "best credit unions in [state]" listicles carry enormous weight in recommendation answers. So do reviews, local press, rate aggregators, and how consistently your name, branches, and details appear everywhere they show up. If your institution reads three different ways across the web, you are a fuzzy entity, and fuzzy entities do not get recommended. They get skipped in favor of the bank that reads clean.
And there is no single scoreboard. ChatGPT retrieves differently than Gemini, which retrieves differently than Perplexity or Claude. You can be described accurately in one and described as a small regional institution with limited products in another, on the same day, from the same underlying facts. Worse, every one of those conversations is private. There is no keyword report. You cannot see what your members are asking or what came back.
So the honest version is this: adding an FAQ page is not going to fix it. Neither is a redesign. This is a different discipline than the one your team has been practicing for fifteen years, and the credit unions treating it as a checklist item are going to spend a year confirming that.
In 1951, a Memphis homebuilder named Kemmons Wilson took his wife, his mother, and five kids on a road trip to Washington, D.C. The motels were filthy, inconsistent, and charged extra for every child. He came home furious and told his wife he was going to build a chain of 400 decent motels where kids stayed free.
He borrowed $325,000 against his house and opened the first Holiday Inn on Summer Avenue in Memphis in August 1952. He franchised the concept in 1957. By 1972 there were more than 1,400 of them and he was on the cover of Time.
Here is the part I care about. Wilson's real insight was not the swimming pool or the free ice. It was location. He built on the outskirts of cities along the highways, and as the Interstate Highway System went in through the 1950s and 60s, his company bought land next to the new interchanges before the traffic arrived.
The thousands of tourist courts sitting on the old U.S. routes did not get worse. Same rooms. Same owners. Same prices. They just ended up on a road nobody drove anymore.
Your credit union is not bad at what it does. Your loan officers are better than the bank's. Your rates are usually better. Your fees are almost always better. None of that matters if you are sitting on the old road.
The traffic pattern is changing again. Right now. The new interchange is an answer box.
I am going to give you the one thing you can do without anyone's help, because I think every CEO in this industry should do it before the next board meeting.
Open ChatGPT. Then Gemini. Then Perplexity. Ask each one the questions your members actually ask. Best credit union in your city. Where should I get an auto loan. Can I join if I live in the next county. Is my money safe at a credit union. Then ask each one, plainly, what it thinks of your institution by name.
Screenshot every answer.
You will find one of three things. You are absent. You are present but described wrong, with stale rates or products you discontinued or a merger that never happened. Or you are lumped in as a generic small local option next to a bank that gets a paragraph of specifics.
All three are marketing problems. None of them are visible in any report your team currently runs. That is why nobody has told you.
"We cannot control what an AI says about us."
Correct. You cannot. You can control what it has to work with.
Right now, for most credit unions, what it has to work with is a rate aggregator, a competitor's blog post, a five-year-old news article, and a PDF. Silence is not a compliance position. It is a decision to let someone else describe you.
And if you are being pitched on this already, and you will be, ask hard questions before you sign anything. Ask what your baseline is across each engine, not just Google. Ask how they will measure being described accurately, not just being mentioned. Ask what happens to the work when the models change again in six months, because they will. A dashboard that reports your "AI visibility score" is not a strategy. Plenty of vendors are about to sell your industry a very expensive one.
Marketing to machines feels undignified after twenty years of building brands for people. I get it. But Wilson did not fall in love with motels. He fell in love with being where the traffic was going.
Go run those fifteen minutes. Bring the screenshots to your leadership team and put them on the wall.
Then decide whether the people who built you a website in 2021 are the right ones to fix what you find. If you want a second opinion on what those answers mean and what it would actually take to change them, that is the conversation I am having with credit unions every week right now. I am happy to have it with you.