STAT's Tom Capper: track brand mentions, not citations, in AI answers
Brands measuring their presence in AI answers should treat mentions and recommendations as the outcome, with cited URLs as supporting evidence rather than the headline metric, according to Tom Capper, Director of Search Product Strategy at STAT. An AI answer can recommend a product without linking to its site, so a report built only on whether a domain earned a citation can miss the recommendation entirely.
Capper used an AI-generated car recommendation as an example: Audi was the brand being recommended while the cited links pointed to other websites. He advised defining the brand entities worth tracking, including sub-brands, and applying the same discipline to competitors so comparisons hold up. Citations still have their uses, he said, such as identifying the review sites and industry publications that shape answers, which teams can then work to improve their representation on.
On the variability of AI answers, Capper framed it as a sampling problem rather than an obstacle, likening it to the location- and device-specific results SEOs already handle. A small prompt set gives an unstable picture, so he recommended realistic questions drawn from sources such as People Also Ask, with personas, locations, budgets and needs in mind. STAT collects answers through the web interface because API responses can differ from what users see.
For connecting visibility to revenue, Capper suggested combining evidence instead of relying on one attribution model. Post-purchase surveys can capture customers who say an AI recommendation helped them discover a brand, and passing those answers into analytics as a dimension lets teams compare self-reported discovery with the channel analytics credits. He also cautioned that sentiment scores can amplify small differences between otherwise neutral answers.
Reported by Search Engine Journal ↗
Topics: Measurement, Citations