Google has given bank and credit union marketers a better way to answer a question that has been difficult to measure: Are our website pages showing up in AI search?
The new reporting is useful. It can help a marketing team identify which pages appear in Google’s AI experiences and decide where to investigate further. But there is a difference between getting a link displayed and getting someone to open an account. Financial institutions need to understand that difference before another impression count becomes the centerpiece of a board presentation.
The bigger opportunity is using this information to examine whether a bank’s website actually helps someone choose it. That means looking beyond blog traffic to the product details, eligibility information, local expertise and application process that support a banking decision.
What Google’s AI search report actually shows
Google announced its Search Generative AI performance reports on June 3, 2026, and says the rollout reached websites worldwide on August 31. There are dedicated reports for Search and Discover. The AI impressions remain included in overall performance reporting, so they should not be added to those totals as extra exposure.
For bank marketers, that last point matters. A new reporting view does not mean the institution suddenly gained a new audience. It makes an existing part of its visibility easier to examine.
The Search report covers AI Overviews and AI Mode. Its metric is impressions: how often links to a site were shown within those features. Available breakdowns include pages, countries, devices and dates.
The current report does not provide a query breakdown, dedicated click counts or click-through rates. Country data also cannot establish visibility within a particular branch market. Google says a missing report usually means a site has not received enough qualifying impressions; exclusion settings are another possible explanation.
Those limitations should shape how the data gets used. It is evidence of exposure within specified Google features, rather than a complete account of AI discovery or customer acquisition.
Start with the pages that matter to your business
A bank could receive substantial AI visibility for a general article about budgeting while its checking, savings and lending pages receive little attention. That article may have value, but its impressions alone do not tell leadership whether the institution is becoming more competitive for new relationships.
Start by sorting visible URLs into a few practical groups:
| Page group | What to investigate |
|---|---|
| Checking, savings and certificate pages | Whether priority deposit products are represented |
| Mortgage, auto and other lending pages | Whether visibility aligns with current lending goals |
| Membership and account-opening information | Whether prospective customers can understand eligibility and requirements |
| Branch and local market pages | Whether the institution provides useful information about the communities it serves |
| Educational articles and calculators | Whether useful research content connects naturally to relevant products |
| Fraud, support and existing-customer resources | Whether exposure primarily supports service rather than acquisition |
This is an analytical recommendation, not a set of categories supplied by Google. The purpose is to give an impression count some business context.
For example, an institution trying to grow small-business deposits should investigate whether its visible content includes business checking, account documentation requirements, treasury services and practical guidance for local business owners. Growth in unrelated consumer education content would not answer the same question.
Review the actual pages, too. A URL can look commercially relevant in a spreadsheet while offering little useful information once someone opens it.
Your product pages may need more work than your blog
Before commissioning another batch of financial education articles, read the pages where people are supposed to choose an account or loan.
A checking page built around “banking that fits your life” leaves basic questions unanswered. What does the account cost? How does someone avoid the monthly fee? Is there a minimum opening deposit? What happens if the account is overdrawn? What must a new customer provide to apply?
A credit union has another question to resolve: Can this person join? “Become a member today” does not explain a field of membership or the steps required to establish membership.
The same problem shows up on lending pages. A prospective borrower needs more than a promise of competitive rates and personal service. Depending on the product, useful information may include available terms, collateral requirements, fees, application documents and the distinction between an advertised starting rate and the rate an individual applicant may receive.
A practical page review should ask:
- Can a visitor identify the product’s costs and major conditions without searching several unrelated pages?
- Are eligibility requirements explained in language a new customer would understand?
- Do advertised rates and offers agree with the linked disclosures?
- Does the page answer the questions employees regularly hear before someone applies?
- Is the next step clear, and does it lead to the correct application?
These improvements are worth making for prospective customers regardless of how often Google displays the page in an AI answer. They also give the marketing team something concrete to improve rather than chasing an abstract visibility score.
Local knowledge gives institutions something useful to publish
A community bank or credit union does not need to imitate a national financial publisher’s entire content operation. It can start with information it knows firsthand and has a reason to explain.
That might include how membership works for employees of a particular organization, what a local business owner needs to open an account, or how a mortgage program available through the institution actually works. The strongest subjects often come from branch employees, loan officers and contact center staff.
Instead of another broad “homebuying tips” article, consider an explanation of a specific borrower question: What documents should a self-employed applicant gather before speaking with a mortgage lender? Where do applicants commonly misunderstand the institution’s process? Which questions require a conversation with a loan officer because the answer depends on the borrower?
Product and lending staff should review those explanations before publication. Their involvement can add substance that a generic article lacks.
The distinction is between having something useful to say and publishing a page because a keyword tool suggested it. A local angle earns its place when it changes the information a reader needs. Replacing a city name in otherwise identical copy adds little.
Do not mistake an SEO sales pitch for a Google requirement
Google’s guidance says its generative AI features rely on core Search ranking and quality systems. Its recommendations continue to emphasize useful original content and sound technical SEO. Eligible pages must be indexed, eligible for a search snippet and included in Search generative AI features; eligibility does not guarantee appearance.
Google also explicitly says special AI text files, mandatory content “chunking” and special schema markup are unnecessary for these features. An llms.txt file does not improve Google Search visibility.
For a financial institution evaluating an AI search proposal, the practical question is what work will actually be completed. An agency should be able to identify the pages it will improve, the technical problems it will investigate, the subject experts it needs and the evidence it will use to assess progress.
Ask for those specifics before approving a new retainer. A service described as “generative engine optimization” still needs a defensible scope, access to the right data and a clear connection to the institution’s priorities.
Third-party monitoring can support that work, but ask how any visibility score is calculated. Which questions are tested? Which platforms? Which locations? How often? A result from a defined sample of prompts should be described as that sample, with its limitations.
Accuracy belongs in the review
For a bank or credit union, appearing in an answer is only part of the concern. The institution should also investigate whether the information being presented is accurate.
Build a repeatable set of manual checks around important products and common questions. Examples might include membership eligibility, monthly checking fees, certificate terms or whether a particular branch offers a service.
These checks are separate research. The Search Console report does not supply the underlying query list, and a manually observed answer should not be presented as the experience of every searcher.
When an answer contains a mistake, record the question, date, platform, response and cited sources. Then investigate what the institution controls. An outdated promotion, conflicting fee information or an unclear eligibility page may warrant correction even if the AI answer changes later.
If the cited information comes from a third party, determine whether that publisher has a correction process. Fixing an owned page cannot guarantee that every system or outside website will immediately reflect the change.
Assign responsibility for maintaining the source information. Rates, fees, product availability and membership rules need an owner and a review process. More exposure increases the reason to keep those details consistent.
Keep AI impressions separate from applications and revenue
The reporting problem gets worse when a marketing team tries to turn visibility into a precise acquisition claim.
Suppose AI impressions increase while checking applications also rise. That is worth investigating, but it does not establish that the impressions caused the applications. A rate promotion, paid campaign, seasonal demand or a better application experience could also explain the change.
Report the evidence in separate layers:
| Measurement | What it helps answer |
|---|---|
| Google AI impressions | Are links to our pages appearing in Google’s covered AI features? |
| Website acquisition and engagement | How are measurable visitors arriving and using the site? |
| Application starts and completions | Are visitors progressing through the application process? |
| Funded accounts, booked loans and balances | What business resulted, where records can be connected appropriately? |
Do not divide all Google organic visits by AI impressions and label the result an AI click-through rate. The numerator and denominator would represent different populations.
Review application measurement as well. If an account-opening or lending platform sits on a separate domain, confirm that the institution can follow the permitted journey across that handoff. A tracking gap can make a productive page look ineffective or conceal where applicants abandon the process.
Where attribution remains incomplete, describe the limitation. Leadership can make a better decision from an honest partial picture than from a precise-looking number built on assumptions.
Review the participation setting before changing it
Google provides a Search generative AI control under Search Console settings. It can include or exclude a site’s links and content from covered generative AI features, including AI Overviews, AI Mode and generative features in Discover. Properties can also inherit a parent property’s setting.
Google says this control does not determine rankings or inclusion in other parts of Search, and it is separate from AI training controls.
For banks and credit unions, the decision should involve the people responsible for marketing, the website and risk review. Document the current configuration and who owns it. If someone proposes a change, define the reason and the information that will be reviewed afterward.
Exclusion should not be treated as an automatic fix for an inaccurate answer. Other sources can still discuss the institution, and the underlying accuracy problem may exist elsewhere.
What the first month should accomplish
A useful first month should produce more than another dashboard.
Begin with a baseline for priority product pages and a review of participation settings. Identify the commercially important pages receiving exposure, then examine their content and application paths. Select a manageable set of high-priority improvements rather than rewriting the entire site.
Next, bring in product experts to resolve missing information and conflicting details. Record publication dates and other campaign changes so later comparisons have context. Compare consistent reporting periods and investigate substantial changes before calling them a trend.
The report to leadership should state which products are represented, what the team changed, what happened in measurable website and application activity, and what remains unknown. If a metric cannot support a claim, leave that claim out.
Google’s new reporting gives financial marketers a useful place to begin. The work that follows still comes down to helping someone understand the institution, choose a suitable product and complete the next step. An impression count can help direct that work; it cannot stand in for the result.
“Banks and credit unions should be asking which products are showing up in AI search and whether those pages give someone a reason to choose them. If your checking page won’t explain the fees and your membership page leaves people guessing about eligibility, fix that before buying another AI marketing package. More impressions won’t solve a website that can’t answer a prospective customer’s basic questions,” said Robert Richardson, CEO of Richardson Marketing.
