Category: High-yield savings · Oʻahu
Platform: MetriFi GEO
Baseline Visibility: 0%
Achieved Visibility: 27.38%
Time to Impact: First measurement window
Outcome: WIN
When someone asks an AI assistant where to find the best certificate of deposit rates in Oʻahu, a familiar set of Hawaii institutions tends to come back. For most of 2025, Pearl Hawaii Federal Credit Union was not one of them, even though its rates were competitive.
Over three sequential experiments, that changed. Visibility across six tracked prompts moved from 0% to 27.38%, and on the single most valuable question (“the best CD rates in Oʻahu”), Pearl Hawaii went from never appearing to appearing in more than seven of every ten answers. (Read what is generative engine optimization and why it matters for credit unions and banks.)
The more useful part of the story is not the number. It is the order in which things happened. The rounds that published new content moved very little. The round that made existing pages more structured and retrievable is where the large gain shows up.
A second pattern runs alongside it, and it may be the more portable one. The gains were not broad. Every question that moved was local and product-specific. Every question that did not move was either definitional or driven purely by price.
What was measured
Six prompts were tracked throughout, unchanged across all three experiments:
• Where in Oʻahu can I find the best savings account options?
• Where in Oʻahu can I find the best rates for Certificates of Deposit?
• Where in Oʻahu, Hawaii can I find the best certificate of deposit (CD) rates?
• Where in Oʻahu, Hawaii can I open a high-yield savings account?
• What’s the difference between a certificate and a certificate of deposit?
• Where can I open a savings account in Oʻahu, Hawaii to earn the most money?
Visibility is the share of tracked AI responses that name the institution. The opening baseline measured 0% across all six prompts. Throughout, the institutions being named instead were consistently Bank of Hawaii, HawaiiUSA Federal Credit Union, American Savings Bank, First Hawaiian Bank, and Hawaii State Federal Credit Union.
Rounds one and two: content earned a first foothold
The first two experiments were content plays: a dedicated guide to CD rates in Oʻahu, and a question-and-answer comparison article positioning the credit union among local options rather than above them.
Together they registered a small first gain. Overall visibility moved from 0% to 2.78%, carried by a single prompt (“Where in Oʻahu can I find the best savings account options?”) appearing for the first time.
One note on how to read these numbers, because two figures in this piece can look like they disagree. Overall visibility is the average across all six tracked prompts. The single prompt that moved reached 16.67% on its own, and one prompt at 16.67% with five at zero averages to 2.78% across six. The results in the round three section work the same way: the six individual results average to the 27.38% overall figure.
We are treating these two rounds as one combined effort rather than two independent wins. The articles were published close together, their measurement windows overlap, and both register the same movement on the same prompt at the same time. When two things ship that close together, it is not fair to credit one over the other, and the honest read is a single modest foothold rather than two stacking results.
What is clear is what did not happen: five of the six prompts, including both head CD terms, did not move at all. The new content coincided with a first appearance on one question, but nothing shifted on the questions that mattered most.
Round three: structure moved the needle
The third experiment published no new article. Instead, Pearl Hawaii worked from a checklist of technical and structural improvements to pages that already existed, the kind of changes that make a page easier for an AI system to read, quote, and trust:
• Structured data (schema) describing the CD products, rates, and terms
• Current APY, minimum deposit, and term shown clearly in text for each CD
• Hero and page copy naming Oʻahu, the product, and current rates
• A “Why choose a CD in Oʻahu from Pearl Hawaii?” section and a linked FAQ
• Plain-language local terms (“Oʻahu CD rates,” “Honolulu certificate of deposit”)
• Branch locations, contact details, and a visible “last updated” date
• About-page metadata and member testimonials
• Page-speed and mobile improvements
After these changes, visibility moved from 2.78% to 27.38% in a single measurement window. Prompt by prompt, before and after:
• “Where in Oʻahu, Hawaii can I find the best certificate of deposit (CD) rates?” 0% before, 71.43% after.
• “Where in Oʻahu can I find the best rates for Certificates of Deposit?” 0% before, 50% after.
• “Where in Oʻahu can I find the best savings account options?” 16.67% before, 28.57% after.
• “Where in Oʻahu, Hawaii can I open a high-yield savings account?” 0% before, 14.29% after.
• “What’s the difference between a certificate and a certificate of deposit?” 0% before, 0% after.
• “Where can I open a savings account in Oʻahu, Hawaii to earn the most money?” 0% before, 0% after.
The two head CD prompts, stuck at 0% through two rounds of publishing, moved to 71.43% and 50%. Four of six prompts improved, and every one that moved was a local, product-specific question.
What we can and can’t attribute
The pattern is the finding: the large gain arrived with the structural round, not with the content rounds. Two articles were followed by a 2.78% foothold on one prompt. A round focused on making existing pages more retrievable was followed by a roughly tenfold increase, concentrated on the most valuable questions.
We want to be straight about the evidence behind that.
The round-three improvements were carried out by the credit union and self-validated against the checklist we provided. We measured a clear visibility gain in the window that followed. What we cannot independently pin down is the exact date each individual change went live: archived records of the site are incomplete, automated inspection does not reliably detect every change, and structured data in particular is easy to miss from the outside. So we can say with confidence that the gain followed the structural work, and that it landed on the questions that work targeted. We are not claiming to isolate which single item on the checklist did the most.
A few other limits are worth naming. The two earlier articles could still be contributing on a lag, since the third round’s baseline sits just downstream of them. Nothing ran against a control. And these are point-in-time measurements over a defined window, not a permanent position.
Where it stands now
That last point is not hypothetical, and it is worth reporting rather than leaving to the imagination.
Re-measuring the same six prompts roughly nine months later, Pearl Hawaii sits at about 16.7% overall: 57% on the head CD question, 29% on the second CD question, 14% on savings account options, and zero on the remaining three. That is a real step down from the 27.38% peak, and still several times the 2.78% the credit union carried into the structural round. This check drew on a smaller sample than the original measurement windows, so read it as direction rather than precision.
The honest summary is that most of the gain persisted and some of it decayed. Visibility earned this way behaves like a position to defend, not a result to bank.
The two prompts that didn’t move
Two prompts stayed at zero, and both are worth understanding.
“What’s the difference between a certificate and a certificate of deposit?” is a definitional question with no local intent. There is little reason for an AI to name a specific institution when answering it, and across every tracked response, none was named: not Pearl Hawaii, and not a competitor.
“Where can I open a savings account in Oʻahu, Hawaii to earn the most money?” is a superlative comparison, and the answers went to whoever posted the highest headline yield at the time. Sometimes that was an online-only bank. More often it was another local institution running a promotional rate. Either way Pearl Hawaii was not competing on headline yield, and no amount of page structure was going to change that. It suggests a pricing constraint more than a content gap.
This is the second pattern, and it is worth stating plainly. Every prompt that moved carried local, product-specific intent. Every prompt that didn’t was either definitional or purely price-driven. The structural work did not lift visibility evenly across the board; it lifted it precisely where the questions matched what the pages were about. Knowing which questions you can realistically win is as useful as winning them, and it is the difference between a tracked prompt set that can improve and one that will read zero no matter what you publish.
What this suggests for other institutions
For a financial institution that has competitive rates but doesn’t appear in AI answers about its own products, this case points to a practical order of operations: before writing more, make sure the pages you already have can be read and quoted. Structured, specific, locally framed pages, with real numbers in plain text and a clear last-updated date, accompanied a far larger move for Pearl Hawaii than two new articles did.
The second lesson is about where to aim. Pick the questions where a local, product-specific answer is the natural one. Definitional questions and pure price comparisons are unlikely to name you however well your pages are built.
It is one institution and one product category, so treat it as a well-measured example rather than a universal rule. But the direction is consistent with what we see repeatedly: when a page is genuinely invisible, the fix is often not more content. It is making what you already have easier to find.
Results were measured in MetriFi GEO using Pearl Hawaii Federal Credit Union’s tracked campaign data, and are published with the credit union’s permission. MetriFi GEO measures how often financial institutions are named in AI-generated answers and runs experiments to test what changes that. To see how your institution currently appears, talk with us.
