The Brainpan AIVI™ Insurance Benchmark
The Brainpan AIVI™ (AI Visibility Index) measures how often and how favorably insurance brands are retrieved, cited, and recommended by AI answer engines. Each quarterly edition scores named insurers on citation share, recommendation rank, and model-by-model coverage across ChatGPT, Gemini, Perplexity, Copilot, and Claude.
Every quarter, re-scored.
The Q3 2026 edition publishes in October 2026. Future editions will publish here as they are released — this page always links to the most recent benchmark so citations and links accumulate to a single, stable URL. Each edition is also published as open JSON: the Q2 2026 edition exposes a Top 10 leaderboard, an OpenAPI description, and a record for every named brand.
The top ten, scored on every component.
These are the ten highest-scoring carriers of the 134 evaluated in the Q2 2026 edition. Every figure below is organic — produced by the engine unprompted, with no paid placement and no brand-name query seeding. Share of model is computed against the study-wide universe of 4,328 organic mentions, not against the ten brands shown here.
Brainpan AIVI™ Insurance Benchmark, Q2 2026. Top 10 of 134 evaluated carriers. Also available as open JSON under CC BY 4.0.
| Rank | Brand | AIVI™ | Mentions | Share of model | Top-3 rate | Recommendation rate | Citations |
|---|---|---|---|---|---|---|---|
| 1 | State Farm | 457 | 10.6% | 79.9% | 40.9% | 129 | |
| 2 | Progressive | 308 | 7.1% | 58.4% | 41.2% | 113 | |
| 3 | USAA | 270 | 6.2% | 56.3% | 46.3% | 103 | |
| 4 | GEICO | 265 | 6.1% | 75.1% | 42.6% | 94 | |
| 5 | Nationwide | 252 | 5.8% | 46.4% | 41.3% | 119 | |
| 6 | Allstate | 276 | 6.4% | 55.8% | 37.7% | 83 | |
| 7 | Travelers | 170 | 3.9% | 59.4% | 34.1% | 77 | |
| 8 | The Hartford | 134 | 3.1% | 88.1% | 27.6% | 57 | |
| 9 | Chubb | 140 | 3.2% | 60.7% | 32.1% | 60 | |
| 10 | Amica | 124 | 2.9% | 66.9% | 39.5% | 55 |
State Farm sets the scale at 100 by holding 10.6% of every organic insurance mention in the study — more than the eighth, ninth, and tenth brands combined. The steepness of that curve is the finding: AI answer surfaces concentrate far harder than search results do, and the gap between first and tenth is a 60-point spread on a normalized index.
What the first edition established.
The Q2 2026 edition evaluated 134 carriers against 300 stratified prompts across five engines, producing 1,500 model responses, 4,328 organic brand mentions, and 1,742 organic citations. Three findings from that run shape how the series is read.
Most AI mentions carry no citation at all. ChatGPT, Gemini, and Claude together produced 2,549 of the 4,328 organic brand mentions in the study and attached a source link to none of them. Perplexity cited on every mention and Copilot on 95.7%. Any insurance marketing program that measures AI visibility through referral traffic or citation tracking alone is blind to the majority of the surface where its brand is actually being discussed.
Volume and recommendation are separate problems. Brands with high mention counts routinely score below brands mentioned less often, because the mentions are references rather than recommendations. The composite is built to expose that gap rather than average it away — the Hartford teardown works through a single carrier that held the best answer-position rate in the study and still finished eighth overall.
Position is a category-level variable, not a brand-level one. Top-three rates cluster tightly by engine — 65.8% on ChatGPT and 65.6% on Copilot — which means placement is largely determined by how an engine composes insurance answers, and the leverage for an individual carrier sits in entering the consideration set at all.
The top-ten leaderboard, the per-engine breakdown, and the machine-readable endpoints are free to read on the Q2 2026 edition page. The full index — all 134 evaluated carriers scored on every component, with the per-engine detail behind each score — is available as a one-time purchase.
Single-organization commercial license / One-time payment — delivered within one business day.Where ranking and recommendation come apart.
The single most useful thing the index does is refuse to average two different problems into one number. A brand can be named constantly and recommended rarely, and the composite is built to expose that rather than smooth it over. Ranked left to right below, the AIVI™ column descends cleanly — and the recommendation column does not follow it.
AIVI™ score and recommendation rate, Q2 2026 AIVI top 10. Panels use different scales; values are direct. USAA is highlighted.
| Brand (AIVI rank) | AIVI™ score | Recommendation rate | Organic mentions |
|---|---|---|---|
| 1. State Farm | 100.0 | 40.9% | 457 |
| 2. Progressive | 71.2 | 41.2% | 308 |
| 3. USAA | 69.1 | 46.3% | 270 |
| 4. GEICO | 65.6 | 42.6% | 265 |
| 5. Nationwide | 63.4 | 41.3% | 252 |
| 6. Allstate | 61.6 | 37.7% | 276 |
| 7. Travelers | 46.0 | 34.1% | 170 |
| 8. The Hartford | 42.0 | 27.6% | 134 |
| 9. Chubb | 41.2 | 32.1% | 140 |
| 10. Amica | 39.9 | 39.5% | 124 |
USAA is the clearest case. It ranks third on the composite with 270 organic mentions, well behind State Farm's 457, but converts a higher share of those mentions into an actual recommendation than any other brand in the top ten — 46.3%. It is winning the mentions it gets. The Hartford is the mirror image: eighth on the composite, the best answer-position rate in the entire study at 88.1%, and the lowest recommendation rate in the top ten at 27.6%. It is present in the answer and losing inside it.
Those are two different diagnoses that call for two different programs, and a single visibility number would have hidden both. The Hartford teardown works one of them through end to end, and the Allstate teardown works the other — third-highest share of model in the study, sixth on the composite, beaten by a carrier named 24 fewer times. The Nationwide teardown takes the carrier that beat it — last in the published top ten on both placement measures, first on citation rate, fifth overall.
One score, five components, no black box.
A brand's AIVI™ score is a weighted composite of five components, normalized so the category leader in each edition scores 100. Every component is measured from organic model output — the brand mentions and citations an engine produces unprompted, with no paid placement and no brand-name query seeding.
- Share of model — 35%. How much of the category's total organic mention volume belongs to the brand.
- Recommendation strength — 25%. How often a mention is an actual recommendation rather than a passing reference. A brand can be named constantly and recommended rarely; this is the component that separates the two.
- Position weighting — 20%. Where the brand lands inside the answer. Being named first in a five-brand answer is not the same result as being named fifth, and the composite does not treat it as one.
- Citation influence — 15%. Whether the engine attaches a source link to the brand, and whose source it is.
- Top-three rate — 5%. How often the brand appears in the first three named positions across the prompt set.
The two components most often misread are share of model, which measures how much of the category conversation a brand holds and not whether it is preferred, and citation influence, which measures whether an engine attaches a source to the brand at all rather than how good that source is.
Citation authority and citation independence are reported alongside the index as companion metrics. They are diagnostic — they explain a score — but they are not weighted into it. The complete controlling document, including prompt-set construction, attribution rules, the stated limitations, and the process any named insurer can use to challenge a published figure, is the versioned AIVI™ Methodology and Correction Policy.
One quarter is a snapshot. Four is a trend line.
The benchmark re-runs every quarter against the same prompt set, the same five engines, and the same scoring method, so editions are directly comparable. That comparability is the point. A single edition tells a carrier where it stands. The second edition tells it whether anything it changed actually moved — which is the question no AI visibility vendor can currently answer with evidence, because almost nobody is measuring the same way twice.
Q2 2026 is live now. The Q3 2026 edition publishes in October 2026 and will carry quarter-over-quarter movement for every scored carrier: who gained share of model, who converted more mentions into recommendations, and where the engines themselves shifted.
There are two ways to buy in.
- The full index, $3,500 one-time. All 134 evaluated carriers scored on every component, with the per-engine detail behind each score. Single-organization commercial license / One-time payment — delivered within one business day.
- Quarterly Benchmark and Strategic Playbook, $12,500 per quarter. The full index every quarter as it publishes, plus quarter-over-quarter movement analysis and a written playbook mapping the movement to the specific content and citation work behind it.
The top-ten leaderboard, the per-engine breakdown, and the machine-readable endpoints stay free to read, permanently, under CC BY 4.0.
Single-organization commercial license / One-time payment — delivered within one business day.Visibility across the engines people actually ask.
- Citation share — how often a brand is cited in AI answers.
- Recommendation rank — where a brand lands when engines recommend.
- Model coverage — presence across ChatGPT, Gemini, Perplexity, Copilot, and Claude.
- Repeatable methodology for quarter-over-quarter comparison.
See your own citation footprint now.
Brands can request an AI Visibility Audit to understand their position against the published leaderboard.