Who answers when your customer asks an AI?
18 studies of the insurance market in France. For each segment, we asked some AI engines 25 buyer questions and recorded the brands they name and the sites they cite.
- insurance segments
- 18
- buyer questions per segment
- 25
- runs per question
- 5
- answers kept per segment
- 125
- answers recorded in total
- 2,250
What we measured
What the AI answers, before your customer compares
A future policyholder now asks an AI first: “health insurance for freelancers”, “cheaper borrower insurance”, “liability insurance for a tradesperson”. The answer names a few brands, cites a few pages, and that is where the shortlist takes shape.
We asked these questions the way a customer would write them, without rewording, in separate conversations. Every answer is kept in full, with its sources. None of the studies rates the quality of the policies: they describe what was answered, on that day.
- 01
Brands named
Which brands appear in the text of the answer, how many times out of 125, and which ones never come up.
- 02
Sites cited
Which pages serve as sources: public sites, comparison sites, media, brand sites. Your listing on a third-party site counts too.
- 03
Answer stability
The same question asked 5 times does not give the same answer 5 times. We measure how often a brand comes back.
From one segment to the next
The share of answers with no brand changes everything
Depending on the segment, the AI readily names brands or answers without citing a single one. Where it names none, the spot is open.
- Borrower insurance85 %
- Income protection for business owners78 %
- Liability for tradespeople and construction78 %
- Group health insurance73 %
- Life insurance71 %
- Car insurance for young drivers57 %
- Health insurance for seniors54 %
- Health insurance for the self‑employed50 %
- Fleet and commercial vehicles50 %
- Cyber insurance for SMEs47 %
- Renters insurance46 %
- Dental and optical cover43 %
- Health insurance40 %
- Ecommerce and cyber for small businesses38 %
- Motorcycles and two-wheelers33 %
- Pet insurance32 %
- Liability for independent and medical professions27 %
- Travel insurance22 %
7 studies out of 18 have a public website as their most cited source.
APRIL is the most named brand in 4 studies.
In 3 studies, the most named brand gets no more than 15 answers out of 125.
Calculated from the 18 studies.
The 18 studies
Pick your segment
Each study details the brand ranking, the sites cited, answer stability, an action plan and the 25 questions asked. Reading time: 10 minutes.
Studies in French
Health and protection
6 studiesBorrower and life
2 studiesMotor, home and personal lines
6 studiesProfessional liability and cyber
4 studiesPanel. 25 questions, 5 runs each, 125 complete answers. Collected on October 27, 2026, in France, through an AI web interface. The exact model is not disclosed.
Mention. The brand name appears in the text of the answer, counted once per answer and per brand. Links alone do not count.
Limits. Non-representative sample. 5 runs do not make a trend. Answers change from one day to the next. No technical site scan. No rating of policy quality.
Frequently asked questions
Reading these studies correctly
The studies refer to “some AI engines”. Answers were collected through an AI web interface, in France, on October 27, 2026. The exact model is not disclosed and the studies do not compare several AI engines.
And your brand: what does the AI answer?
The studies give a snapshot of a segment. The AI visibility audit measures your brand on your own questions, then shows which pages and sources to work on first.