We Asked 4 AI Engines 100 US Insurance Questions. Here's Who Gets Recommended.

When someone asks an AI assistant which insurer to buy, the answer is not pulled from carrier websites. We sent 100 real US insurance buyer questions, across auto, home, renters, life, pet, disability, and small business, to four engines (ChatGPT, Perplexity, Gemini, and Google AI Overviews) with web search on, and recorded which insurers each one named and which sources it cited. The pattern is consistent: a handful of incumbents win almost every answer, the citations come from aggregators and Reddit, and each engine reads a different set of sources.

Method in brief: 100 prompts × 4 engines, US, June 2026. 15 insurers tracked (incumbent, insurtech, and marketplace). Share of voice = the count of the 100 prompts whose answer named that insurer. Full method and caveats are at the end.

Incumbents own the AI answer

The same seven carriers win on every engine. The bar shows average share of voice across the four engines (out of 100 prompts); the columns show each engine.

Incumbent vs insurtech vs marketplace

Share of each engine's insurer mentions, by company type. Google AI Overviews is the only engine that gives marketplaces meaningful share.

The answer is assembled off third-party sites, not carriers

Engines cite roughly 8 to 16 sources per answer, and most are not the carriers themselves. Where the citations come from differs sharply by engine, so the surfaces a brand needs to win are different on each one.

“Other” absorbs long-tail carrier and niche-publisher domains plus a small regulator/industry share.

Where challengers actually break in

Insurtechs and marketplaces are not absent everywhere. They win specific lines: Next Insurance in small business, Lemonade in renters and pet, and the marketplaces in explicit comparison queries. These are the realistic battlegrounds for a challenger brand.

What this means for insurance brands

If your brand is not one of the seven incumbents, an AI assistant will rarely name you unless you have engineered presence on the surfaces it actually reads. Those surfaces are not your own website. They are review and comparison sites, editorial aggregators, Reddit, and YouTube, and the weighting is different on every engine. Winning the AI answer in insurance means showing up accurately on those third-party surfaces, in the specific lines where you can compete, and monitoring what each engine says about your premiums and coverage.

That is the work we do. See how we approach it on the financial services and insurance solution, or run a live audit to see where your brand stands today.

Methodology

We compiled 100 US insurance buyer prompts spanning auto, homeowners, renters, life, pet, disability, umbrella, small business, comparison, and direct brand checks. Each prompt was sent once to four engines with web search enabled: ChatGPT (gpt-4.1), Perplexity (sonar), Gemini (2.5 Flash), and Google AI Overviews, geo-targeted to the United States, in June 2026. We tracked 15 insurers (incumbent carriers, insurtechs, and marketplaces) and recorded which were named in each answer and which source domains each engine cited. Share of voice is the number of the 100 prompts whose answer named a given insurer.

This is a June 2026 snapshot. AI answers are volatile and shift over time and by location, so treat the shares as directional rather than fixed. Google AI Overviews did not appear for 13 of the 100 queries.