NEW MEDIA ADVISORS · ORIGINAL RESEARCH · AUGUST 2026
Who wins AI search in banking
We analyzed over 14,000 AI answers about banks and credit unions across 129 U.S. markets. A clear pattern emerged: the institutions that an AI engine recommends depends far less on size and more on the type of question being asked.
The Banking AI Visibility Report is intended for CMOs, heads of digital, and retail and commercial banking marketing leaders.
U.S. markets tested
institution recommendations analyzed
share of names held by national banks
of citations from sources banks don’t own
THE SHORT VERSION
There is no single “AI visibility” score to manage
Banks have spent years measuring their ranking visibility in Google search. Now, AI engines often sit in front of traditional search, answering questions before the customer ever sees a list of blue links. And AI does not pick who to recommend the same way search engine rankings do.
When someone asks ChatGPT, Gemini, or Perplexity to recommend a bank, the model often runs a live web search, reads the pages it trusts most, and hands back a shortlist. The institutions it names is influenced by the customer’s location, product need, and intent. If you ask a broad, generic banking question then the national banks brands typically surface. If you ask which banks fit a specific person in a specific place for a specific need, then the answers drastically change.
That means there is no one number to chase. There is a portfolio of customer questions, in specific markets and product categories, and the job is to identify the ones your institution can realistically win. Below, you will find the top-level findings of the Banking AI Visibility Report. The full report shows where the openings are, patterns discovered across a variety of local markets, and proposed a path forward for our institution to compete.
Five things senior marketers should take from our findings
- AI is becoming a meaningful part of how banks and credit unions get discovered, and the shift is measurable in consumer behavior.
- The type of question, not institution size, decides which banks and credit unions are most likely to be recommended.
- Local brand and product relevance create real openings for regional banks, community banks, and credit unions.
- Traditional SEO helps your content enter the source pool that AI reads; third-party brand coverage helps shape the AI recommendation.
- There is a market-by-market, product-by-product effect, and the institutions that take action now will have a head start on those that wait.
FINDING 01
The question determines who gets recommended
The single most useful thing we found: when we analyze the same 129 markets through consumer questions versus business questions, and you get opposite leaderboards.
Across consumer categories, credit unions lead the AI recommendation list, taking roughly 61% of recommendations and more than 70% on home equity lines and personal loans once we weighted the results for AI engine usage market share. On those same personal-loan questions, national banks register between 1% and 2% AI visibility.
When we analyzed business and commercial questions, the type of institutions winning completely shifts. Regional and community banks lead, taking 60% to 85% of recommendations, peaking at 85% on acquisition financing and 80% on commercial banking relationships. Credit unions, dominant in consumer questions, fall as low as 5% on treasury services.
Testing the same markets and same engines, we find that brand relevance to the person and the need, not brand size, is driving the outcomes.
Share of the recommendation list flips by segment
6%
5%
Share of recommendation names by institution type, usage-weighted. Source: New Media Advisors analysis (DataForSEO study).
Across consumer categories, credit unions lead the AI recommendation list, taking roughly 61% of recommendations and more than 70% on home equity lines and personal loans once we weighted the results for AI engine usage market share. On those same personal-loan questions, national banks register between 1% and 2% AI visibility.
When we analyzed business and commercial questions, the type of institutions winning completely shifts. Regional and community banks lead, taking 60% to 85% of recommendations, peaking at 85% on acquisition financing and 80% on commercial banking relationships. Credit unions, dominant in consumer questions, fall as low as 5% on treasury services.
Testing the same markets and same engines, we find that brand relevance to the person and the need, not brand size, is driving the outcomes.
Share of the recommendation list flips by segment
6%
5%
Share of recommendation names by institution type, usage-weighted. Source: New Media Advisors analysis (DataForSEO study).
FINDING 02
National banks are 3.7% of the recommended brands, and that is not weakness
Across 6,579 open-ended AI answers, the six banks with true nationwide retail scale account for just 3.7% of all the institutions named once we weight to engine usage by market share. It is an easy number to misread, so let’s break it down.
A low AI visibility share is different from being rarely recommended. At least one national bank shows up in roughly one in four answers: Chase 10% of the time, Bank of America 8% of the time. National banks are present in many local lists. They are simply not the bulk of the recommendation lists, because the other five or six names in answers are regional banks and credit unions.
National brands often shine in engines fewest people use. Bank of America draws about 72% of its mentions from Perplexity, which is roughly two percent of real AI usage by market share. Weighting to real usage pulls the national share down, not up. The generic, non-localized question edge is real, but it lives in broad queries.
The practical takeaway for a regional bank or credit union: your competition for most local answers are other local institutions, not the national brands you may assume you are up against. That reframes where a marketing team should be focused.
Across 6,579 open-ended AI answers, the six banks with true nationwide retail scale account for just 3.7% of all the institutions named once we weight to engine usage by market share. It is an easy number to misread, so let’s break it down.
A low AI visibility share is different from being rarely recommended. At least one national bank shows up in roughly one in four answers: Chase 10% of the time, Bank of America 8% of the time. National banks are present in many local lists. They are simply not the bulk of the recommendation lists, because the other five or six names in answers are regional banks and credit unions.
National brands often shine in engines fewest people use. Bank of America draws about 72% of its mentions from Perplexity, which is roughly two percent of real AI usage by market share. Weighting to real usage pulls the national share down, not up. The generic, non-localized question edge is real, but it lives in broad queries.
The practical takeaway for a regional bank or credit union: your competition for most local answers are other local institutions, not the national brands you may assume you are up against. That reframes where a marketing team should be focused.
Modern SEO is the entry point to AI visibility
The strongest, cleanest relationship in the data is between a domain’s AI citations and its visibility in Google’s top organic results.
Across all the institutions we could match, AI citation counts and Google top three rankings correlate at 0.81, rising to 0.90 among the largest banks. On a scale where 0 means no relationship and 1 means a perfect agreement, 0.81 is strong. Domains that rank well in Google are far more likely to be cited as sources by AI engines.
This is the bridge from the SEO work most banks already manage. If your pages do not rank, they are less likely to be in the set of pages AI models read when performing a query fan-out, and a bank that is not in the pool is unlikely to be named. Being retrieved is often the price of entry, particularly in smaller city markets.
It is an association, not proof of cause, and we are careful about that in the report. But the direction and implications are practical: do not stand up a separate “GEO” program and forsake the SEO you already manage.
For the tactical playbook behind these findings, see our guide to SEO and AI search strategies for bank marketers.
The strongest, cleanest relationship in the data is between a domain’s AI citations and its visibility in Google’s top organic results.
Across all the institutions we could match, AI citation counts and Google top three rankings correlate at 0.81, rising to 0.90 among the largest banks. On a scale where 0 means no relationship and 1 means a perfect agreement, 0.81 is strong. Domains that rank well in Google are far more likely to be cited as sources by AI engines.
This is the bridge from the SEO work most banks already manage. If your pages do not rank, they are less likely to be in the set of pages AI models read when performing a query fan-out, and a bank that is not in the pool is unlikely to be named. Being retrieved is often the price of entry, particularly in smaller city markets.
It is an association, not proof of cause, and we are careful about that in the report. But the direction and implications are practical: do not stand up a separate “GEO” program and forsake the SEO you already manage.
For the tactical playbook behind these findings, see our guide to SEO and AI search strategies for bank marketers.
FINDING 04
Most of what AI retrieves sits on websites you don’t own
Your bank website matters, but it does not control the answer. When we look at the sources AI engines most often retrieve to build an answer, independent third parties are the majority: about 62% of citations on national questions and 52% at the city level. Independent publishers alone, the comparison and review sites, are roughly 42% nationally.
Three brand names show up constantly. NerdWallet is retrieved in about 31% of chats, Bankrate in 24%, and Forbes in 19%. These are the pages the models read before they answer, and they appear across most markets.
Institution-owned websites still matter as they are the citation source 38% of the time nationally and closer to 48% in local markets, where a bank’s own pages carry more weight. Outside research validates the findings: across all industries, Muck Rack finds roughly 94% of web pages cited in AI answers come from third-party sources. Owned content is necessary, but alone it is not sufficient.
Your bank website matters, but it does not control the answer. When we look at the sources AI engines most often retrieve to build an answer, independent third parties are the majority: about 62% of citations on national questions and 52% at the city level. Independent publishers alone, the comparison and review sites, are roughly 42% nationally.
Three brand names show up constantly. NerdWallet is retrieved in about 31% of chats, Bankrate in 24%, and Forbes in 19%. These are the pages the models read before they answer, and they appear across most markets.
Institution-owned websites still matter as they are the citation source 38% of the time nationally and closer to 48% in local markets, where a bank’s own pages carry more weight. Outside research validates the findings: across all industries, Muck Rack finds roughly 94% of web pages cited in AI answers come from third-party sources. Owned content is necessary, but alone it is not sufficient.
FINDING 05
The AI engines don’t agree, so one blended score is misleading
Treating “AI search” as a single channel hides the divergence between the engines, and that divergence matters once you weight for how many people use these platforms.
ChatGPT and Gemini, which together carry the majority of AI usage, lean toward regional and local brands. Perplexity leans most towards the national bank brands, but Perplexity has about only two percent share. So, the national advantage that looks impressive in one engine is concentrated in the one with low usage.
The credit-union results make the point more cleanly. Credit unions take about 60% of consumer recommendations in ChatGPT and 66% in Gemini, the two dominant engines by usage, versus 41% in Perplexity. The local pattern is strongest for the AI engines that are most used, not weakest.
For a marketing team, do not manage AI visibility to a single blended number. Measure recommendation and visibility per engine and prioritize the engines your audiences most often use. A brand that is strong in one engine can be weak in another, and a blended average will hide both.
Credit-union share of consumer recommendations, by engine
Engine usage share from Similarweb, May 2026. Credit-union dominance holds in the two most-used engines. Source: New Media Advisors analysis (DataForSEO study).
Treating “AI search” as a single channel hides the divergence between the engines, and that divergence matters once you weight for how many people use these platforms.
ChatGPT and Gemini, which together carry the majority of AI usage, lean toward regional and local brands. Perplexity leans most towards the national bank brands, but Perplexity has about only two percent share. So, the national advantage that looks impressive in one engine is concentrated in the one with low usage.
The credit-union results make the point more cleanly. Credit unions take about 60% of consumer recommendations in ChatGPT and 66% in Gemini, the two dominant engines by usage, versus 41% in Perplexity. The local pattern is strongest for the AI engines that are most used, not weakest.
For a marketing team, do not manage AI visibility to a single blended number. Measure recommendation and visibility per engine and prioritize the engines your audiences most often use. A brand that is strong in one engine can be weak in another, and a blended average will hide both.
Credit-union share of consumer recommendations, by engine
Engine usage share from Similarweb, May 2026. Credit-union dominance holds in the two most-used engines. Source: New Media Advisors analysis (DataForSEO study).
MARKET BY MARKET
A similar prompt, two places, two different answers
We tracked 13 markets and ran the same prompts multiple times during the study. Whether a market leans local or national comes down to one thing: does it have a dominant home-market bank? A few highlights from the Banking AI Visibility Report.
Bank of America and Chase lead, but Nashville-based Pinnacle Financial Partners and First Horizon tie for third, and Pinnacle holds the strongest average position of any brand in the market. Fully competitive.
WHY IS THIS HAPPENING
AI disrupts the distribution advantage that scale used to win
For decades, acquisition in banking ran on distribution: branch density, ad budget, and brand recall. The biggest banks in a market were often the default answer because it was the most visible one.
AI discovery is leveling the playing field. The customer asks a question and receives a shortlist of the most relevant, best-sourced answers to that exact question. Being the answer now matters more than owning the distribution. That is why a three-branch credit union can win a local lending query, and why a regional bank can own commercial relationships in its market while barely registering on generic national questions.
These dynamics are new. AI is recommending institutions that are specific, locally relevant, and well-represented on the third-party sites the models trust. The models lean less on generic content and do not overweight for brand size. For most banks and credit unions, the next step is not deciding whether AI discovery matters. It is identifying the types of customer questions where they are absent, where the brand is described inaccurately, and where they are losing to weaker competitors.
Where a bank should start
- Identify the questions and topics worth winning. Choose the combinations of market, product, and customer need where you have real presence and a high likelihood to win.
- Measure the baseline. Build a prompt set around those real customer questions and measure recommendation presence (AI visibility) and source presence (citations) across the engines your audiences use. It is important to note where the models inaccurately describe your brand.
- Strengthen the content you own on your website. Make product and local-market pages specific, factual, and crawlable, so they enter the citation consideration set. This is where your existing SEO investment carries over directly.
- Earn the third-party authority. Work with publishers, comparison sites, and local sources the AI models most often cite, because most of what AI reads to build answers is not on your own domain.
- Connect AI visibility share to growth. Track how visibility gains impact branded search demand, and downstream any measurables lift in completed applications, funded accounts, and commercial leads. Some attribution to business impacts will resonate more with leadership than a vanity score.
If you’re ready to build an AI visibility program, learn more about our SEO for banks and credit unions services.
HOW WE DID IT
Two studies, three data sources, with clear caveats
This is original research, not a screenshot of a vendor tracking dashboard. While directional, we built it to be triangulated and to hold up to scrutiny.
open-ended AI answers across 129 markets and 3 engines
7,523
tracked answers on the 200 largest banks, 13 markets, weekly
4,500+
distinct institutions surfaced, including the local long tail
4
engines: ChatGPT, Gemini, Perplexity, Google AI Overview
The open-ended part of the study (leveraging DataforSEO) casts the widest net: 17 realistic prompt scenarios per market, across consumer and business intent, run in all 50 states and 29 metros. That breadth is what surfaces the credit-union and community-bank result that bank-only trackers cannot see. The depth part of the study (tracked in Peec AI) tracks the largest banks week over week to smooth run-to-run variations. Domain-level SEO and citation data comes from Ahrefs. Additional banking AI and SEO data comparison sets were evaluated to look for common patters, via OppAlerts.
We are direct about the limits of the study in the full report. AI answers are probabilistic and vary by run, engine, and location. We interpret the results as directional patterns, not exact forecasts, and we flag thin-sample markets. The study measures which institutions get recommended and which sources get cited. It does not claim to measure account growth or revenue impact, which is the next question marketers are working hard to solve.
WEBINAR REPLAY: Who Wins Financial Discovery in the AI Era?
As AI changes how consumers and businesses find financial institutions, discoverability extends beyond search rankings. In this Financial Brand webinar (July 8, 2026), our co-founders (Scott Gardner and Brent Bouldin) shared insights for how banks and credit unions can stay visible and competitive across search and AI platforms.
Fill out the form below and we’ll send the report right to your inbox.
