AEO for Banks: How Community and Regional Banks Get Named Correctly in AI Answers

Updated October 2026 · Written and maintained by the Progression Agency strategy team

AEO for banks is answer engine optimization for FDIC-insured banks: the work of making a bank’s rates, fees, branches, lenders and insured status readable, checkable and safe to quote, so that ChatGPT, Claude, Gemini, Perplexity, Microsoft Copilot and Google AI Overviews name the bank, and describe it correctly, when someone asks where to keep money or where to borrow it. It is written for community and regional banks with branch networks and deposit and lending products, and for the marketing, compliance and information security staff who share the job. Progression Agency is based in New York City and works with clients across the United States and worldwide.

What is AEO for banks?

AEO for banks is the work that decides what an AI assistant is able to say about a bank, and whether it says it correctly. Three conditions have to hold at once. The bank’s facts exist as text on pages the assistant is allowed to fetch. Each fact agrees with a record the bank does not control, such as its FDIC certificate in BankFind or a loan officer’s NMLS registration. And every sentence a model might lift on its own still meets the advertising rules when it stands alone. The general method of answer engine optimization applies to any business; banking adds a regulator behind almost every sentence.

It is a different job from ranking. SEO for banks on this site covers how people search for a bank, branch pages and Business Profiles, rate page titles and snippets, and what happens to search visibility when branches open, close or merge. This page starts from another reader: a language model assembling one answer from several sources, which may quote a single sentence from a rate page and leave the footnotes, the legend and the link to the account terms behind.

Who this page is written for

• Marketing directors at community banks who have been asked, for the first time, what ChatGPT says about the bank.
• Regional banks with dozens of branches, where one wrong set of hours or one stale CD rate repeats across many answers.
• Compliance officers who review rate, fee and loan copy and now need to know what happens when a sentence is quoted out of its page.
• Information security teams whose bot management decides, often silently, whether AI search crawlers ever see the site.
• Mortgage, small-business and commercial lending teams whose loan officers are looked up by name.
• Banks preparing a merger, a rebrand or a charter conversion, when old names linger in every model’s memory.
• De novo banks that need their insured status and products described correctly from the first week.

What this page leaves to its neighbors

Credit unions have their own page, AEO for credit unions, because NCUA rules and member eligibility change the work. Lenders, insurers and investment firms are covered together in AEO for financial services, and app-first providers in AEO for fintech. Campaigns, paid media and the full advertising rulebook for deposits and loans sit on our bank marketing agency page. What follows is only about being read, checked and quoted by machines.

How many banks compete to be named, and who supervises them?

As of BankFind data refreshed on October 9, 2026, BankFind listed 4,220 active institutions, 4,217 of them FDIC-insured (the other three are noninsured non-deposit companies), and the FDIC flagged 3,790 as community banks. Most are small: 3,171 of the 4,215 that reported a figure held under $1 billion in total assets on June 30, 2026. An assistant asked for a local recommendation is choosing among many such banks, and a bank whose public facts are clear and consistent gives it the least reason to hedge.

The same data shows who supervises each bank, which decides which rule numbers apply to its pages. Among the 4,217 FDIC-insured institutions, the FDIC is the primary federal regulator for 2,643, the Office of the Comptroller of the Currency for 875 and the Federal Reserve for 699; BankFind’s active list also carries three noninsured companies. Anyone can run the same count: the FDIC’s BankFind Suite shows each institution’s certificate, offices and history, and the BankFind API documentation defines every field, including the community bank flag and the primary regulator code.

Why the supervisor matters to a web page

Several rules that bind a bank’s pages are written per regulator. The Equal Housing Lender requirement for dwelling loan advertising sits in 12 CFR 338.3 for FDIC-supervised banks and in 12 CFR 128.4 for federal savings associations under the OCC. The Consumer Financial Protection Bureau supervises banks with more than $10 billion in total assets, which was 160 institutions on the June 30 figures, while the prudential regulator enforces federal consumer financial law for the rest. A compliance reviewer already knows which applies; an agency writing for the bank has to know it too before the first page is drafted.

Branch networks are still large

BankFind counts 76,325 domestic offices operated by active institutions in the 50 states. Of the 78,023 office records in its locations data, 71,878 are coded full-service brick and mortar, 1,382 are drive-through or detached facilities and 821 are loan production offices. Every one of those is a set of hours, services and coordinates that an assistant may be asked about, and a bank with forty branches has forty chances to be quoted wrong.

What do people ask AI assistants about a bank?

Two broad kinds: short factual questions about one named bank, and open requests for a recommendation. The factual ones are the riskiest, because the answer is a single figure or status that the reader may act on. The wording below is illustrative; the pattern is what matters.

Questions about one named bank

These carry the name in the prompt, so the assistant looks for that bank’s site and for records about it. The answer is only as good as what it finds. A routing number published only inside the online banking portal, or hours that differ between the website and the map listing, produce answers that hedge or contradict each other. Fixing them is unglamorous and is usually the first month of the work.

Questions that ask for a shortlist

Requests such as the best bank for a small business in a named town are built from whatever distinguishes one bank from another in text: the industries a business banking page names, the towns its lenders serve, its fee terms, coverage in local news. A bank that describes itself only in slogans gives the model nothing to compare.

Questions after a rate change, a merger or a closure

Models and search indexes refresh on their own schedules. Between a change and the refresh, the old fact is what gets quoted. The defense is a dated page that says what changed, linked from wherever the old fact used to live, and a correction pass through every outside listing the bank controls.

A starting prompt set for a community bank usually mixes these kinds of question, asked in the customer’s words:

• Whether the bank is FDIC insured, by its exact name and by the name customers actually use.
• The bank’s routing number, for each state or legacy bank where more than one exists.
• Branch hours, drive-through hours and ATM access in two or three of its towns.
• The best certificate or savings rate at a local bank in its main market.
• Which bank in a named town suits a small business in a common local industry.
• Who offers first-time buyer mortgages, construction loans or home equity lines in its counties.
• Whether the bank charges overdraft fees, and how much.
• What happened to the bank after its most recent merger, rebrand or branch closure.
• How to move a direct deposit or open an account online.
• Whether deposits above $250,000 are covered, asked the way a business owner would ask it.

Which sources do AI assistants read when the question is about a bank?

Five kinds: the bank’s own website, public registers run by regulators and industry bodies, rate and comparison publishers, news and trade coverage, and reviews, forums and map listings. Which of them an answer draws on varies by assistant and by question, and each source decides for itself which AI crawlers it admits, so the mix any one answer can draw on is narrower than it looks.

The bank’s own site

It is the only source the bank fully controls and the only one that can carry every required term next to every figure. It is also the source most often closed by accident, through a firewall rule, a rate table drawn by JavaScript, or a fee schedule published only as a scanned PDF.

Regulators’ public records

BankFind, the FDIC’s deposit insurance pages and EDIE, the OCC’s HelpWithMyBank site and the CFPB’s complaint database are the records an assistant can use to confirm what a bank says about itself. Their robots.txt files admit AI crawlers, as the table below shows, but much of the data sits behind a search form or a script, and none of it describes a product the way a bank would; it confirms status, history and location.

Rate and comparison publishers

Rate tables and best-of lists are written to be quoted, which makes them natural material for answers about where to put savings. Each publisher sets its own crawler rules, and the rules differ from one AI vendor to the next. A bank listed on one of them should check that its entry is current, because a stale rate there will be repeated alongside the bank’s name.

News, trade press and associations

Local business news, American Banker, the ABA Banking Journal, the ICBA and state bankers’ associations describe mergers, leadership, new branches and lending programs. Coverage is a third-party statement, so it can confirm what a bank says about itself.

Reviews, forums and map listings

Map listings supply hours and addresses; review sites and forums supply opinions. Their crawler rules vary the most: on October 13, 2026 the robots.txt file Reddit served to a plain request refused every crawler, while Trustpilot and the BBB named and admitted most AI search agents.

What did bank-answer sources let AI crawlers read on October 13, 2026?

Most regulators named no AI crawler at all and so left their pages open; the large rate publishers split, with some refusing Anthropic’s or Perplexity’s agents while admitting OpenAI’s. The table reports one plain request for each site’s robots.txt file, tested against a typical content page for each named agent. It shows what each site asked crawlers to do on that day, not what any crawler actually received.

Two cautions apply. A plain request is not a verified crawler, so a 403 shows only that our request was refused; it says nothing certain about what OAI-SearchBot or PerplexityBot would receive. And a robots.txt file can change on any day. The point for a bank is narrower: no outside source can be relied on to carry its facts, so the bank’s own pages have to be complete and open.

What can an assistant confirm about a bank in a public register?

Insured status, the legal name and certificate, branch locations and history, deposit insurance rules, which agency regulates a national bank, and, for larger banks, complaint records. It cannot confirm a rate, a fee or a product feature from any register; those exist only on the bank’s own pages and wherever the bank has supplied them.

Most of these records are built for a person with a search box, not for a crawler. On October 13, 2026 a plain request for the BankFind Suite’s search page, and for an institution’s details page, returned the same 578-byte shell with an empty application container: the records are drawn in by JavaScript, so a crawler that does not run scripts sees none of them. The FDIC also publishes the data through an open API, and the ABA’s routing lookup is a form for one check at a time. The practical consequence is the same in each case: the bank’s own pages should restate the legal name, certificate number and other register facts in plain text, so that an assistant finds the fact and the reader can still verify it at the source.

The CFPB database and the $10 billion line

The CFPB’s complaint database page states that it does not publish complaints referred to other regulators, such as complaints about depository institutions with less than $10 billion in assets. For most community banks, then, the database says nothing at all, and an assistant asked whether customers complain about the bank will lean on reviews and forums instead. The CFPB also notes that the database is not a statistical sample of consumers’ experiences. A bank should not read either presence or absence there as a verdict.

Loan officers and the NMLS identifier

Under the S.A.F.E. Act rules, each employee of a bank who acts as a mortgage loan originator must register with the Nationwide Mortgage Licensing System and Registry and obtain a unique identifier, and 12 CFR 1007.105 requires the bank to make those identifiers available to consumers and the originator to provide it in the first written communication. Printing the ID on each loan officer’s page meets a reader’s need and gives an assistant a number it can tie to a name.

Routing numbers after mergers

The ABA’s lookup warns that new institutions, mergers, name changes and closings keep the list of payment system participants changing. An assistant is most reliable on a routing number that some page states, and a merged bank may keep more than one valid number for a while. Publish every live number in text, say which accounts or states it applies to, and date the page.

How do ChatGPT, Claude, Perplexity, Gemini, Copilot and Google reach a bank’s website?

Through named crawlers and user agents that each vendor documents, and only if both robots.txt and the bank’s edge security let them through. Banks guard their sites against automated traffic for good reasons, which makes the second gate the one to check first.

Each vendor’s own page is the place to check the current names: OpenAI’s crawler documentation, Anthropic’s crawler article, Perplexity’s crawler page and Google’s list of common crawlers. Our free AI crawler access checker takes a pasted robots.txt file and shows which of 21 crawler tokens can reach a given path, which are blocked and which the file never addresses.

Firewalls and bot management at a bank

OpenAI and Perplexity both recommend allowing their published IP ranges as well as the robots.txt token, and Perplexity’s documentation includes configuration notes for Cloudflare and AWS web application firewalls. A bank’s security team may be challenging unfamiliar agents by default, and a challenge page can look to a crawler like an empty or blocked site. The fix is a written decision, owned jointly by marketing and information security, that lists which search agents are verified and allowed on public marketing pages and which stay blocked, with online banking and account opening flows kept out of scope.

The written allow list does not need to be long. It should record:

• Each AI search agent the bank admits, by its documented user-agent token.
• How each is verified, for example against the IP list the vendor publishes for it.
• Which paths are in scope: public marketing, rate, branch, product and help pages only.
• Which paths stay closed: online banking, account opening, applications, internal tools.
• Which training tokens the bank refuses, and who decided.
• When the list was last reviewed, and who reviews it after any vendor changes its crawler names.

Google-Extended and the Gemini app

Google-Extended is not a crawler but a control token. Google’s crawler documentation says it governs whether content may be used to train future Gemini models and for grounding in Gemini Apps, and that it has no effect on inclusion in Google Search. A bank that blocks it to stay out of training also leaves the Gemini app without its pages to ground on. That is a legitimate choice; it should be made knowingly.

Search Console’s generative AI control

Google’s Search generative AI control in Search Console decides whether a site’s links and content can appear in AI Overviews, AI Mode and generative features in Discover. Include is the default, Google says the control reached all websites on August 31, 2026, and it is not used as a ranking signal elsewhere in Search. A bank that wants its rate pages cited should confirm it has not been switched to Exclude.

Snippet controls cut both ways

The robots meta tag documentation says a page-level nosnippet rule also stops the content being used as a direct input for AI Overviews and AI Mode, while the data-nosnippet attribute works on a span, div or section. Applied to a whole rate page, nosnippet removes the page from the answers the bank is trying to enter.

What should a bank publish to be named and linked in AI answers?

A small set of plain, dated pages that answer the questions customers actually put to assistants, each in HTML, each agreeing with the public records, and each written so that any one sentence can be quoted without misleading anyone. None of it is exotic. Most banks already hold every fact; they hold it in PDFs, portals, images and branch binders.

Rate pages written to survive quotation

One page per deposit family, with the annual percentage yield, the date it is accurate as of, the minimum balance to obtain it, the fee statement and, for certificates, the term and the early withdrawal penalty statement, all inside the sentence or the table row that carries the number. The next section shows how.

Fee schedules in HTML

A text version of each consumer and business fee schedule, with its effective date, linked from every product page that refers to it. Keep the signed PDF as the record of what customers received; publish the HTML so it can be read. Overdraft and returned-item fees deserve their own short explainer, because a schedule lists them without saying when they apply.

Branch and ATM pages

One page per office with the street address, lobby hours, drive-through hours where they differ, services offered at that office, ATM access, holiday closures and the staff a customer can ask for. The office should match its BankFind record and its map listing exactly. Our bank marketing page carries the Business Profile checklist for branches; Google’s Business Profile guidelines themselves tell banks to use lobby hours where possible and allow an attached ATM its own profile.

Product pages that compare honestly

Each checking, savings, certificate, home equity, mortgage and business product described in plain words: who it suits, what it costs, what it requires, how to open it. A sentence such as “suits a business that deposits cash daily and writes fewer than 200 checks a month” gives an assistant something to match against a question. A sentence such as “banking that grows with you” does not.

Loan officer and banker pages

A page for each mortgage loan originator, business banker and commercial lender, naming the products they handle, the towns and industries they serve, how to reach them and, for originators, the NMLS identifier. Assistants are asked for names; give them names that can be checked.

The About page as an entity record

The legal name exactly as BankFind shows it, every trade name, the FDIC certificate number, the charter type and primary federal regulator, the holding company, the founding year, the main office address, and dated notes on past mergers and name changes. It is the page an assistant reaches for when the question is whether the bank is real and insured.

Short service answers

Routing numbers, wire cut-off times, funds availability, card activation, holiday hours, how to switch a direct deposit, how to report fraud. Each gets a short, current page in text. These are not acquisition pages, but an assistant that cannot find them on the bank’s site will find someone else’s version.

FAQs that a compliance reviewer has read

Questions with the answer directly underneath remain a clear way to publish a fact, even though Google’s documentation updates record that the FAQ rich result stopped appearing in Search on May 7, 2026. Write each answer so it stands alone, because that is how it will be used.

How should a bank write a rate sentence that an assistant may lift on its own?

Put the required terms inside the same sentence or table row as the number. Regulation DD already asks for most of them whenever an annual percentage yield appears; the AEO step is to stop relying on a footnote, a legend or a separate disclosures page to supply them, because a quote may not carry those along.

The rule text is short. 12 CFR 1030.8 says an advertisement that states a rate of return must state it as an annual percentage yield, may use APY once the full term has appeared, and may show the interest rate only alongside and no more conspicuously. Once the APY appears, the advertisement must state, where applicable, that a variable rate may change after opening, how long the APY is offered or that it is accurate as of a specified date, the minimum balance to obtain it, any larger minimum opening deposit, that fees could reduce earnings and, for time accounts, the term and that a penalty will or may apply for early withdrawal. Our bank marketing page walks through the broadcast and Internet treatment; here the question is what survives extraction.

The figures in the right-hand column are invented to show the shape of a sentence, not offers. Two details from the official interpretation matter for this kind of writing: a statement such as one-, three- and five-year CDs available is not a trigger term, and an as-of date must be recent in relation to how often the medium is published; on a website that changes whenever rates move, the safe practice is to date each figure the day it changes. The same interpretation says an electronic advertisement showing a trigger term must clearly refer the reader to where the additional information begins, for example with a direct link. A self-contained sentence and that link are not alternatives: write both.

• The term annual percentage yield in full at least once, then APY.
• The date the figure is accurate as of, or the period it is offered.
• The minimum balance to obtain it, and each tier’s minimum next to that tier’s yield.
• A larger minimum opening deposit, where there is one.
• That fees could reduce earnings, where maintenance or activity fees apply.
• For a variable rate, that it may change after the account is opened.
• For a certificate, the term and that a penalty will or may be imposed for early withdrawal.
• For a bonus, the APY, the time and balance needed to earn it and when it is paid.

One source for every rate on the site

A rate that appears on the rate page, a product page, a promotion banner and a branch page will eventually disagree with itself. Feed every display from one table, print its date beside each figure, and remove superseded promotional pages rather than leaving them live with an old date.

Rate tables as HTML, not images or scripts

An assistant cannot quote a rate it cannot read. A table drawn into the page by JavaScript after load, or posted as an image, may be invisible to some crawlers. Our free extractable content checker measures the gap between a page’s source HTML and its rendered DOM.

Which loan terms set off more disclosure when an assistant repeats them?

Under Regulation Z, a down payment, a number of payments or repayment period, a payment amount or a finance charge stated in a closed-end credit advertisement each require the down payment, the repayment terms and the annual percentage rate to follow, and home equity plans have triggers of their own. A lifted payment figure is exactly that kind of term, so a loan page should never let a payment, a term or a rate stand alone.

For credit secured by a dwelling, section 1026.24(i) also bars calling a product a government loan program unless it is one, such as an FHA or VA loan, and bars using the word counselor for a for-profit mortgage creditor’s staff. Both are easy to write into a plain-language FAQ answer without noticing. The fix for an AI-facing loan page is the same as for rates: write the representative example as one complete sentence, and put a clear statement of a payment calculator’s assumptions next to its result.

Mortgage and home equity pages

Name the product, the representative loan amount, term, rate, APR and payment together, with the date. Then link to the loan officers who originate it, each with an NMLS identifier. AEO for mortgage brokers covers the non-bank side of the same market.

Small-business and commercial credit

Business lending pages are read by owners asking which bank in their town lends to their kind of company. Name the industries, typical loan sizes, the documents the bank asks for and the people who decide. Fair lending rules still apply to every word, as a later section explains.

Can a bank call an account free when an AI answer may repeat the word?

Only if no maintenance or activity fee can be imposed on it, and the official interpretation treats “fees waived” as a similar term. An assistant asked for free checking near a town may repeat what the page says, so the word has to be accurate before anyone else quotes it.

• Maintenance and activity fees include a fee when a minimum balance is not met, a monthly service fee and per-transaction charges.
• A specific feature can be called free, such as no fee for in-branch withdrawals, if the page does not imply the whole account is free.
• An account free only for a stated period, such as the first year, can say so if the period is stated with it.
• An account can be advertised as free for customers who meet a condition unrelated to the account, such as age, under the interpretation’s own example.
• The word profit may not be used for interest paid on an account.
• Promoting the payment of overdrafts brings the disclosures in 12 CFR 1030.11(b): the fee for each overdraft, the kinds of transactions that can incur it, the time to repay, and when the bank will not pay.

Overdraft explainers need particular care, because they are usually written to reassure. The interpretation lists, as ordinarily misleading, describing an overdraft service only as protection against bounced checks when the bank also charges for overdrafts by ATM, debit card or other electronic transfers. A plain, complete answer is both compliant and the kind an assistant can reuse.

What does the FDIC require around insured-status claims, and how does that shape quotable text?

Advertisements for deposits carry the official advertising statement, digital channels that take deposits carry the official digital sign on set pages, non-deposit products carry no suggestion of FDIC insurance, and no one may misstate how much coverage there is. For AI-facing writing, the last two matter most, because a lifted sentence can blur a deposit and an investment that sat in separate boxes on the page.

Part 328 sets the official advertising statement as “Member of the Federal Deposit Insurance Corporation”, with Member FDIC, Member of FDIC, FDIC-Insured or the FDIC symbol allowed as short forms. Listings in directories are among the advertisements that need no statement. Advertisements relating solely to non-deposit or hybrid products may not carry the statement or anything implying federal deposit insurance, and mixed advertisements must keep it clearly apart from the non-deposit portion. Since the January 2026 amendments, published in the Federal Register and effective March 2, 2026 with compliance required by April 1, 2027, the digital sign must appear clearly, continuously and conspicuously on a deposit-taking website’s homepage, its login page and the page where a consumer first starts opening a deposit account.

Coverage stated exactly

The FDIC’s own wording is coverage of $250,000 per depositor, per FDIC-insured bank, for each account ownership category. Part 328’s misrepresentation rules treat a statement that coverage differs from what the law provides as material. A bank’s insurance explainer should use the FDIC’s words, link to EDIE for anyone with larger balances, and never round, simplify or extend them.

Investment and insurance pages on the bank’s domain

Where the site offers annuities, brokerage or insurance, those pages need non-deposit signage stating that the products are not FDIC insured, are not deposits and may lose value, and no deposit insurance badge. Write that status into the body text as well, so an assistant summarizing the page does not attach the bank’s insured status to a product that has none.

Credit unions, for contrast

Federally insured credit unions follow the NCUA instead. 12 CFR 740.5 requires the NCUA’s official advertising statement, such as “Federally insured by NCUA”, in advertisements including the main Internet page, and 12 CFR 707.8 applies Truth in Savings advertising rules that speak of a dividend rate where a bank’s speak of an interest rate. Regulation DD’s own definition of a depository institution excludes credit unions. AEO for credit unions covers that side.

How do fair lending and Equal Housing Lender rules reach pages written for AI?

Through the words, images and examples on lending pages. Regulation B bars statements, in advertising or otherwise, that would discourage a reasonable person from applying on a prohibited basis, and counts images as statements; the Equal Housing Lender rules require a nondiscrimination statement on dwelling loan advertising, which the logotype and legend satisfy, for the banks they cover.

The current text of 12 CFR 1002.4(b), amended in April 2026, says a creditor must not make an oral or written statement directed at applicants or prospective applicants that it knows or should know would cause a reasonable person to believe it would deny or worsen terms because of a prohibited basis characteristic, and that spoken or written words or visual images such as symbols, photographs or videos count. For FDIC-supervised banks, 12 CFR 338.3 requires dwelling loan advertising to state that loans are made without regard to race, color, religion, national origin, sex, handicap or familial status, which the Equal Housing Lender logotype and legend satisfy in written and visual advertising, and bars words, symbols or models that suggest a discriminatory preference. Federal savings associations follow the OCC’s equivalent in 12 CFR 128.4.

Describe products and places, never the people they would suit

An assistant answering who a mortgage is for will reuse the page’s framing. Describe the loan, the property, the costs and the process. Avoid sentences about the kind of household or neighborhood a product fits, and review photographs and their alt text with the same care as the copy.

The legend in text as well as in the logo

A logo image is invisible to a model reading text. Where a lending page carries the Equal Housing Lender logotype, write the legend in the page’s text as well, so the statement survives in any summary of the page.

Where does UDAAP fit when an assistant summarizes a bank’s product?

It is the backstop under every specific rule. The Dodd-Frank Act makes it unlawful for a covered person to engage in an unfair, deceptive or abusive act or practice, and a page that is technically complete but misleading in its overall impression can still fall under it.

12 U.S.C. 5531 describes an unfair practice as one that causes or is likely to cause substantial injury that consumers cannot reasonably avoid and that is not outweighed by benefits, and an abusive one as one that materially interferes with a consumer’s ability to understand a term or condition, or takes unreasonable advantage of a lack of understanding of a product’s material risks, costs or conditions. Under 12 U.S.C. 5515 the CFPB supervises insured depository institutions with more than $10 billion in assets; for smaller banks the prudential regulator holds exclusive enforcement authority relative to the CFPB. The FTC Act’s own grant of authority to the Federal Trade Commission names banks as an exception.

What that means for writing

Conditions belong next to benefits. If an account’s fee waiver depends on direct deposit, say so in the sentence that mentions the waiver. If a promotional rate ends, say when and what follows. A page written this way gives an assistant nothing misleading to summarize, which is the only part of an AI answer a bank can actually control.

• A fee waiver: name the condition, such as a direct deposit or a minimum daily balance, in the same sentence.
• A promotional rate: give the end date and the rate or range that applies afterward.
• A sign-up bonus: state the deposit, the time it must stay and when the bonus is paid.
• A mobile deposit limit or funds availability rule: say when the money can be used, not only that deposits are easy.
• A business account’s free transaction allowance: give the number and the charge per item beyond it.

Does structured data help a bank get cited in AI search?

It helps a page be read consistently, but it is not required. Google states that structured data is not needed for its generative AI search features and that there is no special schema.org markup to add; its value is agreement between the markup, the visible text and the public record.

• BankOrCreditUnion is the schema.org type for a bank’s office, a sub-type of FinancialService, which carries a feesAndCommissionsSpecification property for the fee schedule.
• Google’s local business documentation asks that each location be marked up as its own LocalBusiness, using the most specific sub-type, and supports departments within a business.
• DepositAccount and LoanOrCredit describe products, and annualPercentageRate is a defined property; Google documents no search feature for them.
• Opening hours, address and telephone in markup should match the branch page text and the map listing word for word.
• Never put a rate, fee or term in markup that is not visible on the page; it breaks Google’s general guidelines and creates a second version of the fact.
• Our schema copy validator checks whether the visible copy actually supports the values in the JSON-LD.

How is a bank website checked for AI readiness?

By testing each gate a fact passes through, in order, and recording what fails: crawler permission, edge security, rendering, document formats, page-level controls and agreement with the registers. For a community bank’s public site the check takes about a working week and produces a short list of fixes in priority order.

1. Read robots.txt for every named AI search agent and every training token, and record the bank’s intended position for each.
2. Pull firewall and CDN logs for the agents’ published IP ranges and look for challenges, blocks and rate limits on marketing pages.
3. Load the rate, fee, branch and product templates with JavaScript disabled and note every figure that disappears.
4. List every fact that exists only in a PDF, an image or behind a login, starting with fee schedules and routing numbers.
5. Check page-level nosnippet, noindex and Search Console’s generative AI control for anything excluding pages the bank wants cited.
6. Compare the legal name, certificate, offices, trade names and loan officer IDs on the site with BankFind and the NMLS registry.
7. Read every rate, loan, fee and insurance sentence through the lift test and mark the ones that need rewriting.
8. Run the baseline prompt set and save every answer with its date before anything changes.

An AEO audit packages these checks for any site; for a bank, the register comparison and the lift test are what make it different.

How are AEO results measured for a bank?

With accuracy first and visibility second. A fixed set of prompts is asked in the same words every month, and each answer is scored for whether the bank is named, whether its rate, fees, hours and insured status are stated correctly, and which pages are linked. Search Console and Bing Webmaster Tools report impressions and citations, not a position within an answer.

The prompt set mixes questions about the bank by name, questions about its products in its towns, and the service questions customers ask most. Our free prompt builder produces a twelve-prompt set and a recording sheet to start from, and AI visibility trackers compared covers the paid tools for running it at volume.

What should a bank do when an assistant states a wrong rate, fee or branch fact?

Correct the fact where the assistant found it, then everywhere else it lives, and log the change so the next prompt run can confirm it. A correction made at the source is the one a bank controls, so the fix always starts there.

1. Save the answer, the date, the assistant and any links it showed.
2. Find the source: the bank’s own page, a stale promotional page, a map listing, a rate publisher’s table or an old news story.
3. Fix the bank’s own page first, with the corrected fact and its date in the sentence.
4. Correct the map listing, the publisher’s table and any directory entry the bank can edit or request a change to.
5. Retire or redirect old promotional pages that still show superseded rates.
6. Re-run the same prompt after the next crawl and record whether the answer changed.

Our guides on why ChatGPT recommends a competitor and why AI visibility drops cover the diagnosis in more depth.

Who at the bank approves content written for AI answers?

Marketing drafts, compliance reviews every rate, fee, loan and insurance sentence, legal handles edge cases, information security owns crawler access, and branch operations and lending confirm the facts they own. An outside agency is a third-party relationship and is managed like one.

The 2023 Interagency Guidance on Third-Party Relationships from the Federal Reserve, the FDIC and the OCC addresses any business arrangement between a banking organization and another entity, names outsourced services and independent consultants among its examples, and states that using a third party does not diminish the bank’s responsibility to operate safely and in compliance with the law. It describes a life cycle of planning, due diligence and selection, contract negotiation, ongoing monitoring and termination. An agency working on a bank’s public pages should expect that process, and should make it easy. The 2023 guidance remains in effect. On September 15, 2026 the OCC, Federal Reserve, FDIC and NCUA proposed to replace it, with comments due November 16, 2026; the proposal keeps the principle that using a third party does not diminish the bank’s responsibility.

A review packet that compliance can work through quickly contains:

• Each changed sentence beside the sentence it replaces.
• The rule each sentence was checked against, by section number.
• The data behind every figure, with its date and owner.
• Screenshots of where the sentence sits on the page, including any table or legend nearby.
• The structured data that repeats any fact in the sentence.
• A note of every outside listing that will need the same change.

What does an AEO engagement for a bank include?

An access and register audit, a rewrite of the pages that carry rates, fees, branches, lenders and insured status, a compliance-ready review packet for each, the technical fixes, and a monthly measurement cycle. The scope grows with the number of branches, products and legacy brands, not with the number of blog posts.

• A crawler and edge-security review with a written allow list for information security.
• A register reconciliation against BankFind, the NMLS registry and the routing lookup.
• Rate and fee pages rebuilt in HTML, with each figure written to pass the lift test.
• Branch and ATM pages that match BankFind and the map listings, with hours and services in text.
• Loan officer and banker pages with NMLS identifiers, products and towns served.
• An About page that works as an entity record: legal name, certificate, charter, regulator, history.
• Service answer pages for routing numbers, cut-off times, funds availability and switching.
• Structured data that matches the visible text, checked page by page.
• A review packet for compliance with every changed sentence and the rule it was checked against.
• A monthly prompt run, accuracy log and report built from Search Console and Bing Webmaster Tools.

What the bank keeps

Every page, every piece of markup, the prompt set, the accuracy log and the review packets stay in the bank’s name. Prompt testing uses public questions typed into public assistants.

How much does AEO for banks cost?

Published US agency price bands for AEO run $1,000 to $4,000 for a one-off audit, $1,500 to $5,000 a month for a small institution, $5,000 to $10,000 for a mid-market one and $10,000 to $20,000 or more at enterprise scale. A Progression AEO retainer is a flat monthly fee on a six-month minimum term, and a quote follows a written scope.

These are US agency market bands, not Progression prices; a Progression retainer carries a six-month minimum term.

What moves the price for a bank

The number of branches and markets, the number of deposit and loan products, legacy brands from past mergers, how long each compliance review cycle takes, whether account opening and rate feeds sit on vendor platforms the bank does not control, and whether the site needs rebuilding rather than editing. The AEO pricing page explains the bands in general.

How do you choose an AEO provider for a bank?

Ask for evidence of understanding the rules, a method that starts with access and records, and a review process your compliance team would accept. Promises of placement in a specific assistant’s answer are a reason to walk away.

Red flags in a proposal

• A fixed number of articles a month with no mention of rate, fee or branch pages.
• An llms.txt file offered as the main deliverable, which Google’s own guide says Google Search can ignore.
• Talk of tricks for chunking content or seeding mentions.
• Any suggestion to relax the firewall for all bots rather than verify specific agents.
• Reports that count mentions but never check whether the facts were right.
• Drafts that reach the website before compliance has seen them.
• A promise that the bank will be the one an assistant recommends.

How to choose an AEO agency covers the general questions.

Answer engine optimization for banks, GEO or AI search for banks: is it one service?

Yes. The names describe the same work for a bank: open access, complete pages, records that agree and sentences that stay compliant when quoted. Google’s AI optimization guide says that, from Google Search’s perspective, optimizing for generative AI search is still search optimization, and that tactics such as chunking content or creating llms.txt files can be ignored for Google Search.

In banking the label matters less than the review. Whatever a provider calls it, the deliverable is the same set of pages, read by the same compliance officer, measured by the same prompts. Our explainers on AEO, GEO and LLM SEO and AEO versus SEO compare the terms for readers who want them.

Where does AEO sit next to bank SEO, paid search and the website?

On the same pages. SEO decides whether a branch or rate page ranks, paid search buys placement for a promotion, and AEO decides whether the page can be read and quoted correctly by an assistant. One rate source, one review process and one set of branch facts serve all three.

• SEO for banks: branch, product and rate pages that rank, and search through mergers and closures.
• Bank marketing agency: deposit and loan campaigns, paid media, email and the full advertising rulebook.
• AEO and paid search together: how the two share data and budgets.
• Local AEO: branch-level visibility in assistants and maps.
• Technical SEO and website speed optimization: the machinery under rate and branch templates.
• ADA-compliant website design: real text in place of images and proper HTML tables serve screen readers and crawlers alike.

Related services

Pages on this site that a bank’s marketing, compliance and digital teams usually read alongside this one.

• Answer engine optimization agency: the full AEO service and method.
• AEO by industry: every market-specific AEO page.
• AEO for credit unions and credit union marketing: the NCUA side of retail banking.
• AEO for financial services and financial services SEO: lenders, insurers and investment firms.
• AEO for fintech, fintech marketing and SEO for fintech: app-first providers and bank partners.
• AEO for mortgage brokers and mortgage marketing: the non-bank lending market.
• AEO for financial advisors and AEO for insurance agencies: advice and coverage businesses that sit beside banks.
• AEO audit and AI visibility audit: a one-off check before a program.
• AEO content writing and SEO content writing: drafting pages that stand up to review.
• How to get cited by AI and how AI search works: the general mechanics.
• How long AEO takes: realistic timelines for any site.
• AI visibility checker and AI visibility tools: measuring where the bank stands.
• Enterprise AEO: for large banks with many brands and states.
• AI answers vs organic search: how the two kinds of traffic differ.
• Generative engine optimization agency and LLM SEO: the same work under other names.
• Google Business Profile optimization and local SEO services: branch listings and maps.
• Review management and online reputation management: what customers say elsewhere.
• Financial advisor marketing and insurance agency marketing: neighboring regulated markets.
• Fintech software development: custom tools, calculators and rate feeds.
• Landing page design and website design and development: rebuilding templates that hide facts.
• International AEO: for banks with customers or branches abroad.
• Bank website design

Frequently asked questions