TL;DR
A benchmark from Brainpan.AI ran 300 consumer-style shopping prompts through ChatGPT, Gemini, Perplexity, Copilot and Claude, yielding 1,500 answers in which 134 insurance brands were named 4,328 times. Its central finding is a gap between showing up in an answer and being endorsed within it. The two moved independently.
Visibility and endorsement are separate metrics
The Hartford illustrates the split. It placed in the top three 88.1% of the time, the strongest showing among the ten highest-ranked brands, yet an active recommendation accompanied only 27.6% of those appearances — the weakest conversion in that group. A brand can be everywhere in the response and still lose at the moment the assistant commits.
Market share turned out to have little predictive power. On home insurance prompts, Amica surfaced in AI answers 10.8 times more often than its filings with US insurance regulators would lead you to expect. Scale in the real market does not reliably buy prominence in the generated one.
Why a US insurance study matters to UK firms
The category is American and so are the carriers, but the mechanism is not. Any UK business whose customers ask an assistant for a recommendation faces the same two-stage funnel, and most are measuring neither stage. Neither search ranking nor advertising budget carries across to either stage with any reliability.
The timing sharpens it. OpenAI confirmed this week that ChatGPT advertising has passed a $1bn annualised run rate, with self-serve buying opening across Europe. A channel where organic prominence already fails to track market share is now also a channel where placement can be purchased — and the study measured the former before the latter arrived at scale in these markets.
Looking forward
The practical instruction is narrow and worth acting on: treat recommendation rate as its own metric, separate from how often a brand simply gets named, because the two do not move together. Comparable efforts are emerging — Conductor has published its own AI search benchmarks, while Evident Insights’ index measures something different again, namely insurers’ internal AI capability rather than how assistants portray them. For UK marketers, the discipline of separating those questions is likely to matter more than any single league table.