AI Agent Trust & Governance
How can credit unions measure their AI visibility?
8 min read
AI engines are already answering questions about credit unions. They are also choosing which sources to cite. That makes AI visibility a governance issue, not a vanity metric. Measure it by tracking mention rate, owned citation rate, third-party citation rate, and citation accuracy across ChatGPT, Perplexity, Google AI Overviews, and Gemini.
Quick answer
The fastest way to measure credit union AI visibility is to run a fixed set of member-intent prompts, score every answer against verified ground truth, and track how often AI mentions your credit union, cites your owned sources, or cites third-party aggregators.
In Senso’s live benchmark across 80 credit unions, the current pattern is clear. Mention rate is about 14%. Owned citation rate is about 13%. Third-party citation rate is about 87%. Total citations tracked are 182,000+.
The metrics that matter
AI visibility is not one number. It is a set of signals.
| Metric | What it measures | Why it matters |
|---|---|---|
| Mention rate | How often AI names your credit union in an answer | Shows whether the model knows you exist |
| Owned citation rate | How often AI cites your credit union’s own sources | Shows how much of the narrative you control |
| Third-party citation rate | How often AI cites Reddit, NerdWallet, Bankrate, Wikipedia, or other outside sources | Shows whether outsiders are shaping the answer |
| Citation accuracy | Whether the citation matches verified ground truth | Shows whether the answer is grounded and audit-ready |
| Response quality | Whether the answer is current, complete, and policy-consistent | Shows whether the response can be used safely |
| Share of voice | Your share of named or cited answers versus peers | Shows your position in the category |
For credit unions, owned citation rate and citation accuracy matter most. A high mention rate does not help if AI points to an outdated fee page or a third-party aggregator.
How to measure AI visibility step by step
1. Compile your verified ground truth
Start with your source of record.
Ingest raw sources for products, rates, policies, branch details, membership rules, and compliance language. Then compile them into a governed, version-controlled knowledge base.
Use one compiled knowledge base for both internal agents and external AI-answer representation. That removes duplication and keeps the answer surface consistent.
2. Build a prompt set from real member questions
Use the questions people actually ask.
Good prompt categories include:
- Membership eligibility
- Loan rates and terms
- Deposit fees
- Branch hours and ATM access
- Fraud and dispute policies
- Business banking options
- Digital banking support
- Credit union mission and field of membership
Keep the prompts stable over time. If the prompts change every week, the benchmark loses value.
3. Run the same prompts across the same models
Measure the same set in:
- ChatGPT
- Perplexity
- Google AI Overviews
- Gemini
Use the same prompt wording each time. Record the date, model, answer text, and citations.
This gives you a repeatable benchmark. It also shows how each model treats your sources differently.
4. Score every answer against verified ground truth
For each response, mark:
- Did the model mention the credit union?
- Did it cite an owned source?
- Did it cite a third-party source?
- Was the answer grounded in verified ground truth?
- Was the answer current?
- Was the answer complete enough for a member to act on?
This is where governance starts. If a CISO or compliance officer asks whether the model cited a current policy, you need a record that proves it.
5. Track the source mix
AI visibility is not just about being mentioned. It is about where the model sends the user next.
In Senso’s credit union benchmark, third-party citations dominate. The top cited domains include:
| Domain | Citations |
|---|---|
| reddit.com | 1,247 |
| forbes.com | 1,187 |
| wikipedia.org | 1,165 |
| nerdwallet.com | 1,058 |
| bankrate.com | 950 |
That pattern tells you something important. If your own sources are not citable, the model fills the gap with someone else’s content.
6. Compare against peers
A single credit union score is useful. A peer benchmark is better.
Use a panel of similar credit unions by size, market, and product mix. Then compare:
- Mention rate
- Owned citation rate
- Third-party citation rate
- Citation accuracy
- Share of voice
This shows whether a problem is internal, category-wide, or model-specific.
7. Assign owners and retest
Every gap should route to an owner.
If a rate page is stale, send it to marketing or product. If a policy answer is wrong, send it to compliance. If a branch answer is incomplete, send it to operations.
Then rerun the benchmark. Measurement only matters if it changes the answer.
What the current benchmark shows
The live credit union benchmark from Senso tracks 80 credit unions across ChatGPT, Perplexity, Google AI Overviews, and Gemini.
The headline numbers are:
- Credit unions tracked: 80
- Mention rate: ~14%
- Owned citation rate: ~13%
- Third-party citation rate: ~87%
- Total citations tracked: 182,000+
That gap is the point.
If AI does not cite your own sources, your institution is present in the question but absent in the answer. For a credit union, that is a visibility problem and an audit problem.
How to read the results
If mention rate is low
The model is not seeing your credit union as a relevant source.
Check whether your public content is easy to compile into a structured, agent-readable format. Check whether your products, policies, and member-facing context are explicit and current.
If owned citation rate is low
The model knows you exist, but it prefers other sources.
That usually means your sources are hard to find, hard to parse, or not strong enough to cite against the member question.
If third-party citation rate is high
Outside sites are shaping your story.
That is common in financial services. It is also fixable. The answer is to make your own sources more complete, more current, and easier for models to cite.
If citation accuracy is low
The model is grounding answers in the wrong place.
This is the main compliance risk. It means the answer may sound right while citing stale or incomplete material.
If response quality is low
The answer is drifting from approved language.
That creates member confusion, extra wait time, and more manual correction work for staff.
What good measurement unlocks
When credit unions measure AI visibility well, three things improve.
First, narrative control improves. In one tracked result, Senso reached 60% narrative control in 4 weeks.
Second, share of voice can move. Senso has seen 0% to 31% share of voice in 90 days.
Third, response quality can rise. Senso has seen 90%+ response quality and a 5x reduction in wait times.
Those numbers matter because they show that visibility is measurable. It is not guesswork.
A simple measurement cadence
A practical cadence looks like this:
- Weekly for active content changes
- Monthly for stable source sets
- Quarterly for peer benchmarking
- Immediately after major policy, rate, or product changes
Credit unions that move faster on policy updates and public content usually see fewer stale answers.
FAQ
What is the best first metric for credit unions?
Start with owned citation rate.
If AI cites your own sources, you have a better chance of controlling the answer. If it cites third parties instead, the problem is usually source readiness or source visibility.
Is mention rate enough?
No.
Mention rate only shows whether the model knows your credit union exists. It does not show whether the answer is grounded, current, or compliant.
What should be measured for regulated teams?
Measure citation accuracy, owned citation rate, source freshness, and answer consistency.
Those four signals matter more than raw volume.
How often should AI visibility be measured?
At least monthly.
If your rates, policies, or product pages change often, measure weekly.
Can this be measured without integration?
Yes.
Senso AI Discovery gives marketing and compliance teams a no-integration audit of public AI responses. It scores those responses for accuracy, brand visibility, and compliance against verified ground truth.
Where to start
If you want a baseline, start with a free audit.
Ingest your raw sources. Compile your credit union’s verified ground truth. Run a fixed prompt set across the major models. Then score the answers against that standard.
That gives you a real AI visibility baseline. It also shows exactly where AI is speaking for your credit union today, and whether it is getting the story right.