STAT's Tom Capper: Measure AI Mentions, Not Just Citations

An AI answer can recommend a product without sending a single click. Tom Capper, Director of Search Product Strategy at STAT, called that the core measurement problem marketers now face in the webinar "Tactical Solutions For The Biggest AI Search Measurement Challenges."

The customer who sees a recommendation may later search the brand, visit directly, or buy offline. Analytics cannot connect that purchase to the earlier answer. Capper's position was direct: zero-click search is an attribution challenge, not simply a traffic problem. His answer was to measure visibility alongside business outcomes rather than expect clicks to describe the customer journey.

Why should brands track mentions separately from citations?

Capper used an AI-generated car recommendation as an example. Audi was the brand recommended. The cited links pointed to other websites. A report checking only whether Audi's domain earned a citation would miss the recommendation itself.

The practical implication: define the brand entities you track, including sub-brands, and apply the same discipline to competitors. Capper drew the line himself:

"So yeah, citations have URLs. That's still valuable tactical data, but I don't think it should be your KPI. Mentions have entities."

The mention is the outcome to understand. The cited URLs explain which sources shape it.

How do you build a representative prompt sample?

AI answers vary for the same prompt, but Capper treated that variability as a sampling problem, comparing it to the location- and device-specific results SEOs already track. A small prompt set gives an unstable picture, especially when a business already monitors a larger keyword set.

He cautioned against feeding a keyword list into a prompt tracker. Instead, he suggested People Also Ask questions as a starting point, then reasoning about the personas, locations, budgets, and needs behind them. He also noted STAT used the web interface to collect answers, because API responses could differ from what users actually saw.

What can citations still tell you?

Citations should not headline the KPI report, but they make campaign work concrete. Capper described starting with pages cited when competitors were recommended. Those pages reveal third-party sources worth investigating, such as review sites or industry publications. Teams can then improve their representation on those sources and monitor whether recommendations shift.

The useful question is not whether citation counts rose, but which sources changed, what work the team did, and how tracked answers changed afterward.

Which visibility metrics belong in the report?

Capper recommended leading indicators that support, not replace, traditional reporting:

• Presence: whether the brand appears in the relevant search result or AI answer.
• Recommendation and context: whether the answer recommends the brand, merely names it, or ties it to a strength or problem.
• SERP prominence: in traditional search, how far down the page a result sits and how much visible space it occupies.

The last distinction matters even when rank holds steady. Capper showed examples where a lower-ranked result appeared above the fold while a higher-ranked result fell below it.

He also warned against sentiment scores. A chart swinging sharply among positive, neutral, and negative may amplify small differences in neutral answers. Qualitative themes behind a mention can prove more useful than a volatile summary score.

Can surveys close the attribution gap?

During the Q&A, Loren Baker asked how teams could connect visibility to the customer journey when multi-touch attribution cannot observe the earlier exposure. Capper advised combining evidence rather than trusting one model. Post-purchase surveys can surface customers who say an AI recommendation helped them discover or choose a brand, while campaign tracking shows whether recommendations increased.

He suggested passing survey answers into analytics as a dimension, so teams can compare self-reported discovery with the credited channel. He also warned that displaying survey options in the same order biases responses toward the first option. This is a fuller account of influence, not a promise to reconstruct every path.

What changes in the next reporting cycle?

Capper's plan comes down to four moves:

• Audit what the dashboard misses: visits and citations measured, mentions and recommendations not.
• Document brand entities, competitors, prompt contexts, and collection environment before comparing over time.
• Record cited pages investigated or influenced, then track how those sources and recommendations change.
• Compare visibility trends with revenue, conversions, and survey answers rather than presenting visibility as a substitute for commercial results.

The next session in the series, "AI Shopping: Get Your Products Found in Google & ChatGPT," features Chris Sachs, VP of Value Enablement, and Aris Vrakas, Product Lead, at seoClarity, on how AI agents decide which products shoppers consider.