A regional fintech serving credit unions, 14 states

The challenge

The client provides core banking and digital channels software to mid-sized credit unions across 14 states. Their buyers are technology committees and credit union CEOs, and the buying cycle is long, conservative, and reference-heavy. The marketing problem was specific: in 2025 their two largest competitors started showing up in LLM responses for category prompts and they did not. The internal team had run the prompts themselves and confirmed it. Claude was particularly stark, citing one incumbent in nearly every relevant response and never naming the client even when the prompt was asking for alternatives. The category is small enough that the buying committee absolutely uses LLMs for vendor shortlisting, and the head of marketing had been told by two prospects that the brand never came up when they asked. The team had a healthy content library but most of it was gated, and trust signals lived inside PDFs that LLMs could not see.

What we shipped

The strategy here was about trust signals more than volume. We mapped the citation sources Claude was leaning on for the category and found a clear pattern: NCUA-related .gov pages, two trade associations, and a small set of credit union trade publications. The client had relationships with most of these but had never been treated as a primary source.

We did six things. First, we ungated the technical documentation library. About 340 pages moved from gated PDF to indexable HTML with proper schema. Second, we built an entity-grade About and Leadership section with verified credentials, board memberships, and a sameAs graph for the executive team linking to NCUA committee work and trade association profiles. Third, we placed nine bylined articles in three credit union trade publications across Q1 and Q2, focusing on regulatory and core-conversion topics where the client had unique data. Fourth, we got the client cited in two NCUA-adjacent industry reports by providing original anonymized data on digital channel adoption. Fifth, we built llms-full.txt at 920KB covering the product, the integration list, the security posture, and the case studies. Sixth, we added structured Article and Organization markup site-wide and submitted to the three vendor directories that credit union buyers are known to use.

We deliberately did not chase volume content. The category does not reward it and the buyers do not trust it. Every page shipped had a named author with verifiable credentials.

The numbers

The RFP volume change was the headline for the client's CFO, but the more interesting shift was in how the RFPs read. Pre-engagement, roughly half of inbound RFPs were already shaped around an incumbent's feature set, meaning the client was being invited to a process they were unlikely to win. By month five, the RFPs were arriving with language that matched the client's positioning, including specific phrases that appeared in the trade-pub bylines and the ungated documentation. Sales win rate on inbound RFPs moved from 18 percent to 27 percent over the engagement. Sales cycle did not shorten meaningfully, which is normal in this category and not something we expected to fix in seven months. Branded search lifted noticeably and the head of marketing started seeing the brand named in committee meetings she sat in on. Two of the larger inbound RFPs came from credit unions that had previously rejected the client in earlier evaluations, and in both cases the technology committee cited the trade-pub bylines as part of why the client was being reconsidered.

What we'd do differently

We were slow to push for the ungating decision. The legal team had real concerns about competitive intelligence leaking, and we accepted that pushback for three weeks before pulling the executive sponsor in. Once the CEO weighed in, the decision took an afternoon. We should have escalated in week one. We also overestimated how much the LLM-citation work would compound from on-domain changes alone. The trade-pub placements were doing most of the heavy lifting by month four, and we could have been more aggressive earlier with the press strategy. We were too cautious about pitch volume in the first six weeks because we were worried about the brand looking opportunistic.

What's next

The engagement renewed for another twelve months with an expanded scope. The next phase is twofold. First, we are building a state-level citation strategy because each state has its own credit union league and trade press, and current LLM responses to "best core banking platform in [state]" are a wide-open opportunity. Second, we are extending the entity work to product managers and engineers, not just executives, because Claude in particular weights individual contributor expertise more than other models. The goal is to push Claude citation share past 60 percent on category prompts and to capture meaningful share on the 14 state-specific query variants by year end. A separate workstream is building a structured data product on credit union digital channel adoption that the client can publish quarterly, on the working hypothesis that recurring original data is one of the most durable citation sources available.