How to Measure a Bank's AI Search Visibility

Key takeaways

• One screenshot of one AI answer proves little, because answers change from run to run and from one platform to the next. Measurement needs a fixed set of customer questions, asked repeatedly on each platform.
• The useful indicators are how often the bank is named, how often it is linked, how accurately it is described and which sources shape the answer, read by product and platform.
• Tools can run the questions at scale, but no single tool covers every platform the same way, and none of the major ones lists a banking-specific feature. The bank still has to define the questions and the metrics.

A board member asks the CMO whether the bank shows up when business owners ask ChatGPT where to open a small business checking account. Someone on the team types the question, takes a screenshot of an answer that names the bank, and the question seems settled. A week later a colleague asks the same question and gets a different list.

Each screenshot shows a real answer, but one answer is not a measurement. AI answers vary from one run to the next, differ by platform, and change when the customer mentions a place or a need. The bank needs to know how often it appears across many answers, on each platform, for each product it cares about, and whether what those answers say about it is correct.

How should banks measure visibility in ChatGPT, Perplexity and Google AI Overviews?

Start from the questions customers ask, and keep them fixed. For each priority product, write the questions a customer would ask at each stage of choosing, from understanding the options to comparing providers and checking a specific bank before applying. Phrase them the way customers do, including the versions that name a city, a type of business or a life situation, since answers change when those details appear. Once the set is agreed, do not edit questions in place. Add or retire them with a date, so month-to-month changes reflect the answers rather than the questions.

Run the set on each platform separately. An index of banks built by 5W, a communications firm, covered 31,500 prompts across five engines and 75 institutions, and found Google AI Overviews the most volatile of them week to week. Its author, the firm's founder, recommends tracking results monthly across engines, so report each platform on its own.

Run each question more than once. Research by SparkToro with Gumshoe, an AI visibility vendor, repeated 12 recommendation prompts nearly 3,000 times and found that the same list of recommendations came back in fewer than 1 in 100 runs. How often a brand appeared across many runs was more stable than where it ranked in any single answer. The study used no banking questions, but the lesson carries over. Report visibility as a rate across repeated runs, and do not read much into position within one answer.

Record the fields in the table below for every answer.

The bank's own analytics add a second, partial view. Google counts traffic from AI Overviews and AI Mode in the Search Console Performance report, within the Web search type, and since August 2026 a separate Generative AI performance report also shows impressions of a site's pages in those features by page, country and device. Bing Webmaster Tools' AI Performance report shows how often a site's pages are cited in Copilot answers. Google Analytics 4 now has a default AI Assistant channel for visits from assistants such as ChatGPT, Gemini and Copilot, which excludes Google's own AI Overviews and AI Mode. OpenAI says ChatGPT adds utm_source=chatgpt.com to referral links for sites that allow its search crawler. Perplexity documents that its answers link to pages, but not how those visits identify themselves. None of these numbers captures an answer that shaped a customer's choice without a click, which is why the answers themselves remain the primary measure.

What KPIs should banks use to track AI search visibility and share of voice?

A small set of indicators, each defined in writing and read by product and platform, works better than a long dashboard or a single score for the whole bank. Tools define citations differently. Ahrefs counts a citation when an answer links to the brand's pages, while Peec AI counts a source citation when the brand's content is used even if the brand is not named. A bank that compares numbers across tools or vendors without fixing the definitions first will be comparing different things.

Link share is the indicator to read against mention rate. When it is low for questions about the bank itself, the gap usually points to the bank's own pages. For comparison questions, it points to the outside sources.

Accuracy deserves a place next to visibility, and it is easy to leave out. An answer that names the bank with last year's rate, or with a fee the account no longer charges, should go to the fix list rather than the visibility report. It is also the indicator the compliance team will care about most, because it shows where outdated or wrong information about the bank's products is circulating.

For leadership, report mention rate, share of voice, link share and accuracy by product and platform each month. Report business outcomes each quarter once there is enough data to assess them.

AI visibility tools and agencies for banks

Tools and agencies do different jobs, and a bank may need a tool, an agency or simply a well-run internal process. A tool runs a set of questions on a schedule and stores the answers. An agency or internal team decides which questions matter, interprets the answers for each product, and works on pages and outside sources to improve those results. That is why the question set should be settled before choosing a tool.

Use the following questions to evaluate whether a tool meets the bank's needs.

• Which platforms and modes does it cover, and is coverage the same for every platform?
• Does it collect answers through official APIs or from the consumer interface? Conductor, for example, states that it uses official APIs, and API answers may differ from what a customer sees in the app.
• Can the bank load its own question set, including location and product variations, and set how often it runs?
• How does it define a mention and a citation? Does it distinguish the bank's own pages from third-party pages that describe the bank?
• Can results be exported for the bank's own reporting?
• What security assurances does it offer, and what happens to the question sets and data the bank loads into it?

Agencies are worth judging on the same basis as any other measurement partner. Ask to see the question set they would use for your products, how they would define each indicator, and an example of the monthly readout, before discussing anything else.

Best tools to track a bank's visibility in ChatGPT and Perplexity

There is no single best tool for every bank, and none of the major tools lists a banking-specific tracking feature. Conductor has a finance industry page, but it describes general compliance and security capabilities rather than banking-specific AI tracking. The table below lists options in alphabetical order, with what each maker says it tracks. It is a starting point for evaluation and not a ranking.

Product pages checked on 30 September 2026. Coverage and features change often, so confirm them in a trial with the bank's own questions before committing.

For ChatGPT and Perplexity specifically, the choice matters less than the setup. Load the same question set into the selected tool and confirm how it collects answers from ChatGPT and Perplexity. In the first month, manually check a sample of questions to see whether the tool's answers match what a customer sees.

Where CB/I Digital fits

Measurement is where CB/I Digital usually begins its work with a bank. We build the question set around its priority products and markets, run it on each AI platform every month, and report mentions, links, accuracy and sources by product. The readout identifies gaps in the bank's own pages, outside sources that need attention, or products that do not yet compete.

The SEO & AI Search services we offer include the monthly measurement and the fixes that follow. If your team is setting up AI visibility reporting, send us your question set and we will review it with you.

Frequently asked questions

How many questions should a bank's measurement set include?

Enough to cover each priority product at each stage of choosing, and few enough to review every month. Start narrow with the products that matter most and widen the set once the monthly readout is running.

Can a bank measure AI visibility without buying a tool?

Yes, at a small scale. A team can run a short question set by hand on each platform, record the same fields every time and repeat it monthly. The effort grows quickly with more products, markets and repeated runs, which is usually the point at which a tool is worth pricing.

How long before the numbers mean something?

Treat the first month as a baseline rather than a result. Running the set weekly for the first four weeks shows how much the answers move on their own, which tells the team how large a change has to be before it counts. After that, a monthly run is usually enough, with a change in the question set recorded whenever the bank launches or retires a product.