Digital Marketing in Banking When Buyers Ask AI

The short version

When a customer asks an AI assistant which bank to use, three things can appear on the same screen: a labeled ad, a link to your product page, and a generated answer that names institutions directly. Digital marketing in banking used to mean winning the first two. Now it also means being readable enough to be named in the third. None of these layers replaces the others, and each one follows different rules.

The ad: bought, labeled, and separate

The top layer is the one you already control with budget. Google states that its search results are independent of its advertising programs, and that ads are labeled and may appear in several locations around free search results. Two practical consequences follow.

First, buying ads does not buy a mention in the organic results or in an AI answer. The ad is labeled as an ad. Second, your paid campaigns and your eligibility to be cited are separate workstreams with separate owners. A bank can run strong campaigns and still be invisible in the generated answer, because the answer is built from pages, not from spend.

The page: still the raw material

The middle layer is your own site, and it matters more than most bank marketers assume. Google's documentation on its AI features says that existing SEO best practices remain relevant for AI Overviews and AI Mode, with no additional requirements or special optimizations. It also says no special AI text files and no special structured data are required.

What is required is basic eligibility. To be eligible as a supporting link, a page must be indexed and eligible for Google Search with a snippet, and eligibility does not guarantee crawling, indexing, or serving. That is a low bar, and many bank sites trip on it anyway, for ordinary reasons:

Product facts locked inside PDF rate sheets or images, where they are hard to quote.

Pages with no effective date, so an engine cannot tell whether the rate is current.

One generic "checking accounts" page covering five products, so no single page answers a specific question.

Key pages blocked, orphaned, or missing a snippet-eligible description.

These are the same weaknesses that hurt classic search. The difference is the cost. In a list of ten blue links, a weak page might still get a click. An answer engine does not hand back a list to browse. It names a few institutions, and a page it cannot read or date is a page it cannot cite.

The answer: a third layer above both

The newest layer is the generated answer itself. Google describes AI Overviews as summaries of complicated topics with links for further exploration, and AI Mode as supporting nuanced questions and complex comparisons with links to supporting websites. Two details from that documentation matter for planning.

AI Overviews do not trigger on every search. Some customer questions produce an answer, others produce a classic results page, and the mix can change. And both features may issue multiple related searches across subtopics and data sources, so their responses and links can vary. A question like "best checking account for a small business in Ohio" can fan out into sub-questions about fees, minimums, and branch access before the answer is assembled.

This is also why you should not assume the answer always sits above the ads, or that it appears at all. Placement varies by query and over time. The safe planning assumption is that for a growing set of comparison questions, a customer may read a generated shortlist before they ever see your page or your ad.

What to check this week

You do not need new tooling to start. You need one real customer question and an hour.

Pick a question your customers actually ask, in their words. "Which bank near me has free business checking" is better than a keyword from a rank tracker.

Ask it in Google and in at least one other assistant. Note whether an AI answer appears, which institutions it names, and which sources it links.

Open the linked sources. Ask what those pages have that yours does not. Usually it is a specific fact, stated in plain text, with a date.

Check your own product page for the same question. Is the fee, the minimum balance, and the eligibility rule written as text, or trapped in a PDF? Is there an effective date? Does one page answer one question?

Confirm the page is indexed and showing a snippet in Search Console. Google includes traffic from AI features in overall Search Console search traffic, so your baseline measurement is already there.

Treat this as advice from observation, not as measured results. The engines change, and what you see today is a snapshot, not a trend line.

What to change, in order

Start with facts, not keywords. Put rates, fees, minimums, and eligibility in plain HTML text on the product page, each with an effective date. Regulated content like rates and disclosures deserves a careful eye before it ships, and your own review process applies, but the format change itself is editorial, not technical.

Then split compound pages. One page per product, per question, gives an engine something specific to cite. A community bank with a dozen well-defined product pages is easier to quote than a larger institution with one page covering everything.

Skip the rumored shortcuts. Google says no special files or markup are required for its AI features, and its requirements should not be generalized to other engines, which publish less about how they select sources. The durable investment across all of them is the same: clear, dated, crawlable facts.

How to tell whether it worked

Re-ask your test question on a schedule, monthly is reasonable, and log three things: whether an answer appeared, whether you were named, and which sources were linked. Watch Search Console for movement on the pages you fixed. When checking by hand runs out, Scanley measures whether the AI engines name your institution across the questions your customers ask.

No change guarantees a citation. What you can guarantee is that your facts are readable, current, and on a page an engine is allowed to use. That is the part of digital marketing in banking that did not exist as a discipline a few years ago, and it is the part you fully control.

Questions people ask