Which insurers do AI assistants actually name?
We asked four AI assistants the questions a buyer would ask, and recorded which carriers they named. The assistants disagree with each other so severely that any single combined figure would be misleading, so every number here carries its engine and its cohort.
What was measured
US Insurance AI Discovery Index — Edition 1. 88 organisations, captured June 2026, methodology version 2026.1.
US regional and mid-market insurance carriers (50 health, 38 commercial P&C) with an owned website and active digital presence, selected before measurement in June 2026. National giants are tracked only as competitor mentions, not ranked members.
The cohort was fixed before measurement began. It divides into 50 regional health plans and 38 regional p&c carriers. These two groups behave so differently that they are never combined into a single figure on this page.
Results by assistant
This is the primary table, and it is deliberately first. Reading a single combined number for “AI” would hide the finding: Google Gemini never named 4.5% of the cohort while Perplexity never named 64.8% of the same carriers, answering the same questions in the same period.
| Assistant | Carriers measured | Never named | Average visibility |
|---|---|---|---|
| ChatGPT | 88 | 43 of 88 (48.9%) | 14.8% |
| Claude | 88 | 43 of 88 (48.9%) | 16.1% |
| Google Gemini | 88 | 4 of 88 (4.5%) | 82.2% |
| Perplexity | 88 | 57 of 88 (64.8%) | 9.1% |
By assistant and cohort
Splitting each assistant by cohort is where the useful detail sits. A carrier reading only the headline for its own segment would draw a very different conclusion from one reading the combined figure.
ChatGPT
| Cohort | Measured | Never named | Average visibility |
|---|---|---|---|
| Regional health plans | 50 | 10 (20%) | 24% |
| Regional P&C carriers | 38 | 33 (86.8%) | 2.6% |
Claude
| Cohort | Measured | Never named | Average visibility |
|---|---|---|---|
| Regional health plans | 50 | 26 (52%) | 13.3% |
| Regional P&C carriers | 38 | 17 (44.7%) | 19.7% |
Google Gemini
| Cohort | Measured | Never named | Average visibility |
|---|---|---|---|
| Regional health plans | 50 | 1 (2%) | 89% |
| Regional P&C carriers | 38 | 3 (7.9%) | 73.2% |
Perplexity
| Cohort | Measured | Never named | Average visibility |
|---|---|---|---|
| Regional health plans | 50 | 19 (38%) | 16% |
| Regional P&C carriers | 38 | 38 (100%) | 0% |
Disclosures
These are reproduced verbatim from the measurement pipeline rather than written for this page. Every figure above is generated from the verified export; none of it is typed by hand.
- Benchmark metric computed on: chatgpt, claude, gemini, perplexity.
- Headline visibility counts a question as covered when at least one measured engine names the organization; per-engine distributions are reported separately because engines differ substantially.
- Canonical discovery questions: p1, p10, p12, p14, p4, p7.
- Capture window: 2026-06-01 through 2026-06-30 (inclusive).
- 419 answer(s) captured before methodology stamping are attributed to methodology 2026.1 by operator assertion.
Answers vary between runs, so presence is reported as a share across repeated runs rather than as a position. See how Veribix measures and what a consideration set is.