Picture a customer who has kept a checking account at the same bank for years and now wants a high-yield savings account. Before visiting any bank's site, they ask ChatGPT where to open one. EMARKETER's AI Visibility Index for financial services, which tracks ChatGPT, suggests how that answer often goes. Fintechs held far higher mention rates within their categories than traditional institutions held in theirs, and the name ChatGPT gave most often for bank accounts was Ally, an online bank.
If the answer lists other providers, the customer's own bank may lose the chance to compete for that deposit before the customer reaches its website. The bank's analytics will not show that the comparison took place.
Large banks can compete for these moments. A fintech with one product publishes about little else, so it often has the most detailed answers in its category. A large bank can point to its own charter, deposit insurance and a long record under supervision, which many newer fintechs cannot yet show. These matter on YMYL topics, as the trust section below explains. That advantage helps only where the bank's product information is complete and easy to find.
It is tempting to explain EMARKETER's numbers with better apps. That explanation does not account for categories where the bank's product is just as good. A more useful place to look is what each kind of company publishes, because that is what AI systems read.
When ChatGPT answers a savings question with search turned on, it typically searches the web, pulls passages from sources it trusts and assembles them into a shortlist. Google's AI Overviews and AI Mode draw on the same index as Search. Google publishes no list of what earns a citation. The table below is our reading of Google's guidance and the research cited in this article.
| What AI systems tend to reward | Why fintechs tend to have it | Where large banks often fall short |
|---|---|---|
| Depth on the topic of the question | Most of what a fintech publishes is about one product, so each page adds context to it | Coverage is spread across hundreds of products, and a priority account may have a single page |
| A passage that answers the question on its own | Pages are written around the customer's question | Rates, fees and eligibility sit in separate tables and disclosure documents |
| Outside sources that confirm the picture | Comparison sites, reviews and press cover the one product closely | Outside coverage describes the bank as a whole, and product details drift between listings |
| Trust on Your Money or Your Life (YMYL) topics | Varies by company, and harder to show with a short track record | The bank's strongest asset, yet licensing, deposit insurance and credentialed staff rarely appear on the product page |
| Pages a crawler can read | Lean sites built recently | Key terms can sit in PDFs or depend on scripts to load, which makes them harder for crawlers to read |
A savings account shows how this plays out. The bank's account may pay a competitive rate, but the rate sits in a table on one page, the fees in a schedule on another and the eligibility rules in a PDF disclosure. A fintech's page for its one account usually answers all three in a few paragraphs, and that is the passage an AI system can quote.
On trust, banks start with more to show. Google's Search Quality Rater Guidelines class financial topics as "Your Money or Your Life" and ask raters to hold those pages to a higher standard of expertise and trustworthiness. Raters have no say over rankings. The guidelines still describe what Google's systems aim to reward, and in our view the same standard shapes which financial sources AI answers rely on. Banks should make their licensing, deposit insurance information and relevant expertise easy to find on product pages, where AI systems read, instead of leaving them in a footer or on an About page.
EMARKETER's later edition of the index reported that ChatGPT recommends financial brands more often when they have broad product coverage, strong visibility across trusted sources and comprehensive content on their own websites. A large bank already has the breadth. The harder requirement is comprehensive content, because the bank needs to provide complete information for each of its hundreds of products.
McKinsey's research on AI search found that top brands in major categories, credit cards among them, can be absent from some AI answers, and that some brands hold a lower share in AI search than their market share and traditional search performance would suggest. McKinsey's finding means market position alone does not secure a place in the answer.
Few large banks can improve every product page at once, and trying usually spreads the budget and the review workload too thin to make a difference anywhere. It works better to decide category by category where to compete, fix those products first, and put extra effort into the questions where a bank has the advantage.
This is a product decision as much as a content one, so marketing and product should make it together. For each category, put the bank's offer next to the fintechs that AI systems already name, and be honest about the comparison.
| Situation in a product category | What the bank should do |
|---|---|
| Fintechs own the category and the bank's product is weaker on what customers ask about, such as fees or app features | Leave visibility here for later. Content cannot fix a product gap, and a page that overstates the product creates compliance risk |
| The product competes, but information about it is scattered or inconsistent | This is where content work is most within the bank's control. Build one complete source and make outside descriptions match it |
| The customer's question depends on scale, a physical network or a relationship | Lead here. These are questions a fintech cannot answer with the same authority |
The middle row is usually the place to start. The product is already competitive. The bank can gather its terms in one place, update outside listings and answer the questions customers ask.
For each priority product, build a clear main page and supporting articles that answer the questions customers ask when comparing providers. The main page should explain who the product suits and who it does not. It should state the costs, the current rate and terms, the eligibility requirements and how to open it. It should also explain how the product compares with the options the customer is weighing. Supporting articles answer the follow-up questions and link back to the main page.
AI systems usually take a passage from a page rather than reading it end to end, so each main page needs a paragraph that answers the core question on its own. We recommend drafting that answer first, then checking that its rates, fees and eligibility terms match the supporting disclosures. Put rates and fees in a text table rather than an image. Name the reviewer, such as a product specialist, and show when the page was last reviewed. On YMYL topics like these, that information helps readers and AI systems judge whether the page is a reliable source. Structured data and a check with JavaScript switched off help crawlers read the page, though neither adds information the page lacks.
The information also has to stay current. An outdated rate is a real risk, because an AI system may repeat it to a customer. Each priority page needs an owner who updates it when terms change.
The bank's own site is only part of what AI systems read. On broad financial questions, a large share of what they cite comes from publishers, comparison sites and other third parties, so those sources need the same current facts. That means current product details in comparison and review listings, app store descriptions that match the product pages, and newsroom material that describes each product the way the product page does.
This work is slow because it crosses teams. Partnerships or affiliate teams manage the comparison listings, communications owns the press record, and product owns the terms. Our article on strategy, ownership and operating model sets out how to divide that work.
Some valuable questions depend on things a large bank has built over decades. A small business that deposits cash every day needs to know where it can do that. A family buying a home may want a loan officer they can meet. A growing company may want treasury services, lending and payroll from one relationship.
A fintech is a weak source on these questions, however good its app. Clear, specific pages on branch and ATM access, cash handling for business accounts, relationship lending and business services give AI systems a reason to name the banks that offer them.
Customers can also ask ChatGPT whether their money is safe with a particular bank. If the bank's site has no page that explains deposit insurance, licensing and how accounts are protected in plain language, the answer gets built from press coverage and forums. A page like that is worth doing early, with its deposit insurance wording reviewed by compliance like any other content. A bank with a long history under supervision can answer this question directly.
Once the gap is clear, it is tempting to copy the fintech playbook wholesale. The part worth copying is concentration on a small number of products, and the table above already explains why weaker categories can wait. Other habits carry more risk for a regulated bank.
Marketing language that runs ahead of the disclosures is one of them. Fintech-style claims about speed or simplicity have to match the account terms, and claims that could mislead a customer carry UDAAP risk, so compliance reviews them before publishing.
Splitting the bank's presence into many small brands to look more focused usually weakens it. Each new name starts with no history, no outside coverage and no trust record, which are the signals AI answers appear to draw on.
Measure each priority category on each AI platform, tracking whether the bank is named, linked and described accurately next to the fintechs and comparison sites in the same answers. Categories where the product already competes are the likeliest to show early change, since the fixes there are within the bank's control, though no timing is certain. Our guide to measuring a bank's AI search visibility sets out the method.
CB/I Digital has been building AI search expertise since 2019, and banking sits at the core of our enterprise AI search practice. We organize the work around a framework we call TRUST. It checks each priority product page for topical authority, relevance to revenue, the customer's journey, signals of trust, and a technical base that AI systems can read. With a bank's marketing and product teams, we help choose the categories worth competing in, build the product pages and supporting content, align the outside sources that describe each product, and report results by category and platform.
If you are weighing where your bank should compete first, our SEO & AI Search services cover that work, and we can talk through how we would approach your categories.
Should a bank publish comparisons with competitors' products?
Comparison content can help customers, but every claim about another provider has to be accurate and current when it is published, and compliance should review it like any other comparative claim. Many banks prefer to describe their own product completely and leave head-to-head tables to third-party sites.
How often should a bank revisit which categories it competes in?
Revisit the choice when a product, its pricing or a competitor's offer changes in a way that shifts the comparison, and review it against the category results each quarter. A category that was not worth chasing last year can become winnable after a product change.
Do reviews and app ratings affect whether a bank is recommended?
They are part of the outside record AI systems can read, and they matter most on the trust question. A customer asking whether a bank is reliable may get an answer drawn partly from reviews and forums. No single review site decides the answer, but a pattern of complaints left unanswered can shape how an AI system describes the bank.
Copyright © 2026 CB/I DIGITAL INC. All rights reserved | Privacy Policy