What seems to drive carrier placement in LLMs.
The instrumentation layer for Pillar 1 (GEO) of the two-pillar LLM distribution model. If you’re reading this, you’ve picked GEO as your first investment — this page tells you what content shapes are actually moving placement.
Time-stability check (2026-04-29 vs 2026-04-26) → 12 of 15 retested observations stable on Perplexity, 3 carrier flips (cheapest, most-reliable, fintech-cyber).
A running observation log, not a study. Each entry is a structured record with the actual per-LLM run arrays, the modal counts, the carriers observed, and links to the reference pages LLMs cite when they return the winning carrier. Available as JSON at /api/observations.json.
Labels: Seems to have impact = at least 2 LLMs return the same modal carrier with count ≥ 3/5. Small / conditional impact = stable in some cases, divergent in others. No observed impact = no LLM returns a stable modal at the threshold. Labels are derived from the run data, not editorial.
Test summary — 11 levers tested
Each row groups observations by lever and adds an actionable recommendation. Numbers in the “finding” column come directly from the per-LLM run arrays in the supporting observations — click a lever to see the full data trail.
| Lever | Intent | Impact | Finding | Recommendation |
|---|---|---|---|---|
| Property-type naming | home | Seems to have impact | Chubb owns luxury + historic at 5/5 + 5/5; Nationwide owns condo at 3/5 + 5/5. Property type is its own retrieval lever. | Each property type behaves like a niche. Carriers should claim one editorial property-type surface and defend it. |
| Emerging-category presence | auto | Seems to have impact | Travelers wins EV-insurance queries 4/5 + 5/5. Tesla Insurance, Root, Lemonade do not place first. | Open emerging categories are claimable by whoever publishes editorial first. Generalist who got there before insurtechs takes the surface. |
| Specialty use-case ownership (inverted) | auto | Seems to have impact | Rideshare → State Farm 3/5 + 5/5. SR-22 → GEICO 5/5 + 3/5. The actual market specialists (Progressive, Dairyland) do not place first. | Brand-specialty alone does not win specialty queries. Specialists must close the editorial-depth gap with comparison-site presence. |
| Regional monopoly (line-specific) | home · auto | Seems to have impact | NJM owns NJ auto 4/5 + 4/5. MAPFRE owns MA home 5/5 + 3/5. But MAPFRE does NOT own MA auto, and NJM does NOT own NJ home — line-of-business dominance is required. | Regional dominance only transfers per line. Audit your top-3 states' citation share by line before assuming carry-over from one product to another. |
| Specialty peril (saturation-dependent) | home | Small / conditional impact | Chubb wins jewelry 5/5 + 5/5 because high-net-worth has saturated editorial coverage. Wildfire and flood do NOT replicate cleanly — generalists win those by default. | Specialty peril is conditional on whether the peril category itself has consumer-editorial saturation. The category-depth check matters more than peril-specificity. |
| Comparison-site citation share | home | Seems to have impact | ~58% of unique Perplexity citation hosts across 8 US cities are consumer-comparison sites (NerdWallet, The Zebra, Bankrate, Policygenius, Insurify). Every winning carrier appears on at least one of these sites in its winning category. | Comparison-site presence is table-stakes for any LLM citation strategy. Verify category-by-category coverage; missing the consumer-comparison surface caps your retrieval ceiling. |
| Price/quality anchor wording | auto | Seems to have impact | Adjective in the query meaningfully shifts the modal. 'cheapest' Seattle → State Farm 5/5 + 5/5 (not GEICO). 'most reliable' → USAA 5/5 + 4/5. 'top-rated' → USAA 4/5 + 3/5. 'best value' and 'premium' show cross-LLM divergence. | Plan editorial content for each price/quality adjective separately. Don't assume 'cheapest' routes to direct writers — the editorial surface answers it differently than the actual market does. |
| Bundled-policy prompt (own retrieval surface) | home | Seems to have impact | Bundle queries do NOT inherit from single-line winners. Boston bundle → Farmers 5/5 + 5/5 (not MAPFRE). LA bundle → AAA 5/5 + 4/5 (not Mercury). Newark NJ bundle → divergent (NJM does not extend its NJ-auto monopoly to bundles). | Bundle prompts route to national bundlers (Farmers, AAA, State Farm). Single-line champions need a separate bundle-content investment if they want to win that query class. |
| Commercial cyber by vertical (B2B — vertical-specific, drift-prone) | commercial | Small / conditional impact | Round 2 (2026-04-26): Chubb won all five cyber verticals on at least one LLM. 2026-05-04 retest: Coalition wins fintech (5/5 + 3/5), healthcare-SMB (3/5 + 3/5), and SaaS (Gemini 5/5). Chubb still wins law-firm (4/5 + 5/5) and manufacturer (Chubb 5/5 on Gemini). The 'B2B inherits B2C pattern' claim shifted in 8 days. | Treat the 'specialist loses to generalist' rule as vertical-specific in commercial cyber, not universal. Coalition's vertical-specific editorial recovers fintech and healthcare-SMB; Chubb's brand-incumbency holds law-firm and manufacturer. See full analysis in the Coalition-rising essay. |
| HQ-city presence | home | No observed impact | Tested across 5 HQ cities. Only 1 of 10 observations matched the predicted HQ'd carrier. Boston (Liberty Mutual HQ) → MAPFRE on both LLMs. Columbus (Nationwide HQ) → Cincinnati / Erie. HQ address has no observed lift. | Drop the HQ-city assumption. Carrier headquarters does not surface in geo queries — regional volume share or editorial presence does. |
| Market specialist without editorial depth | auto | No observed impact | Hagerty (largest US classic-car insurer) does not win classic-car. Dairyland (largest SR-22 monoline writer) does not win SR-22. Tesla Insurance does not win the Tesla query on Perplexity. Specialist brand alone does not surface. | Inverse of the editorial-depth observations. Specialists can only recover their surface by closing the editorial-coverage gap — brand-specialty positioning does not surface in LLM queries. |
Seems to have impact (17)
Property-type naming — luxury home
Chubb wins luxury-home queries on both LLMs at the maximum stability we measured. Cross-LLM agreement at 5/5 + 5/5; under the null this is ~1 in 39 billion.
Property-type naming — historic home
Chubb wins historic-home queries 4/5 on Perplexity (one State Farm slip) and 5/5 on Gemini. Cross-LLM agreement on the same carrier confirms the property-type ownership pattern extends beyond luxury.
Chubb's win across luxury + historic suggests it owns the broader 'premium / elevated-risk property' category, not one specific property type.
Property-type naming — condo
Nationwide wins condo-insurance queries cross-LLM. Perplexity returns Nationwide 3/5 (with two State Farm slips), Gemini returns Nationwide 5/5. Cross-LLM agreement at the modal carrier.
Emerging-category presence — EV insurance
Travelers wins electric-vehicle auto queries 4/5 on Perplexity and 5/5 on Gemini. Neither EV-native insurtechs (Root, Lemonade) nor Tesla Insurance ranks first. Travelers captured EV editorial coverage early.
Specialty use-case ownership — inverted (rideshare)
State Farm wins rideshare-driver queries 3/5 on Perplexity and 5/5 on Gemini. Progressive — the actual market specialist (TNC endorsement pioneer) — does not place first. Editorial depth, not market-specialty status, drives retrieval.
Specialty carriers can recover their surface by closing the editorial-depth gap — brand-specialty alone is not enough.
Specialty use-case ownership — inverted (SR-22)
GEICO wins SR-22 queries 5/5 on Perplexity and 3/5 on Gemini. Dairyland — the largest US SR-22 monoline writer — does not place first. Same editorial-depth pattern as rideshare.
Regional monopoly — NJM owns NJ auto
NJM wins Newark, NJ auto queries 4/5 on both LLMs (one Plymouth Rock per LLM). NJM writes ~1 in 4 NJ auto policies — the cleanest regional-champion case in the dataset.
Caveat: same carrier doesn't win across lines. MAPFRE owns Boston home (5/5 + 3/5) but does NOT own Boston auto (Plymouth Rock + State Farm split). Line-of-business dominance is required.
Regional monopoly — MAPFRE owns MA home
MAPFRE wins Boston, MA home queries 5/5 on Perplexity and 3/5 on Gemini (with Amica slips). The HQ'd carrier (Liberty Mutual, Boston) does not place first — regional volume share matters, not HQ address.
Comparison sites are the dominant Perplexity citation surface
Across 8 US cities × Perplexity in the multi-city probe, ~58% of unique citation hosts are consumer-comparison sites: NerdWallet, The Zebra, Bankrate, Policygenius, Insurify. Every winning carrier in the observations above is named on at least one of these sites in the matching category.
Price anchor — 'cheapest'
perplexity returns State Farm 5/5; gemini returns State Farm 5/5. Verdict: predicted GEICO, but State Farm actually placed first.
Price anchor — 'top-rated'
perplexity returns USAA 4/5; gemini returns USAA 3/5. Verdict: prediction held: USAA placed first.
Price anchor — 'most reliable'
perplexity returns USAA 5/5; gemini returns USAA 4/5. Verdict: predicted State Farm, but USAA actually placed first.
Bundle prompt — Boston, MA (MAPFRE territory)
perplexity returns Farmers 5/5; gemini returns Farmers 5/5. Verdict: predicted MAPFRE, but Farmers actually placed first.
Bundle prompt — Los Angeles, CA
perplexity returns AAA 5/5; gemini returns AAA 4/5. Verdict: predicted Mercury, but AAA actually placed first.
Commercial cyber — mid-market law firm
perplexity returns Chubb 5/5; gemini returns Chubb 5/5. Verdict: predicted Beazley, but Chubb actually placed first.
Commercial cyber — healthcare SMB
perplexity returns Chubb 5/5; gemini returns Chubb 3/5. Verdict: prediction held: Chubb placed first.
Commercial cyber — manufacturer (200 employees)
perplexity returns Chubb 5/5; gemini returns Chubb 4/5. Verdict: predicted Travelers, but Chubb actually placed first.
Small / conditional impact (8)
Specialty-peril coverage — high-value jewelry
Chubb wins jewelry-coverage queries 5/5 on both LLMs. Specialty-peril surfacing replicates cleanly when the category itself has saturated editorial coverage (high-net-worth = Chubb's editorial home turf).
Wildfire and flood specialty perils do NOT replicate at this level — generalists win those by default. Specialty wins are conditional on editorial saturation.
Price anchor — 'best value' (control)
perplexity returns USAA 4/5; gemini returns State Farm 4/5. Verdict: prediction held: USAA placed first.
Price anchor — 'premium'
perplexity returns USAA 4/5; gemini returns Chubb 4/5. Verdict: prediction held: Chubb placed first.
Bundle prompt — Seattle (control)
perplexity returns State Farm 5/5; gemini returns PEMCO 3/5. Verdict: prediction held: State Farm placed first.
Bundle prompt — Newark, NJ (NJM territory)
perplexity returns Farmers 2/5; gemini returns NJM 3/5. Verdict: prediction held: NJM placed first.
Bundle prompt — Atlanta, GA
perplexity returns Progressive 4/5; gemini returns State Farm 2/5. Verdict: predicted Auto-Owners, but Progressive actually placed first.
Commercial cyber — SaaS startup (50 employees)
perplexity returns Chubb 5/5; gemini returns Hiscox 2/5. Verdict: predicted Coalition, but Chubb actually placed first.
Commercial cyber — fintech startup
perplexity returns Chubb 4/5; gemini returns Coalition 3/5. Verdict: predicted At-Bay, but Chubb / Coalition actually placed first.
No observed impact (2)
Being HQ'd in the query city
Tested across 5 HQ cities. Only 1 of 10 observations matched the predicted HQ-city carrier. Bloomington IL → State Farm only on Gemini (Perplexity returned Allstate). Columbus OH (Nationwide HQ) → Cincinnati and Erie. Boston MA (Liberty Mutual HQ) → MAPFRE on both LLMs. HQ address has no observed lift.
Being the actual market specialist — without matching editorial depth
Hagerty (largest US classic-car insurer) does not win the LLM classic-car query. Dairyland (largest SR-22 monoline writer) does not win SR-22. Progressive (rideshare-insurance category pioneer) does not win rideshare. Tesla Insurance does not win the Tesla query on Perplexity. Specialist brand alone does not surface; editorial depth wins.
Inverse of the editorial-depth observations above. Specialists can recover their surface by closing the editorial-coverage gap, not by leaning on category-specialty positioning.
How these observations are made
Each entry comes from an ad-hoc probe: a buyer-shaped query run against two LLMs (Perplexity Sonar Pro + Gemini 2.5 Pro, both with web-search / grounding enabled). Stable queries get 5 runs per LLM. The first-named US insurance carrier in each response becomes one element of the LLM’s run array. The modal carrier across the 5 runs is the stability measure. Impact label derives mechanically from the cross-LLM agreement on that modal.
Deeper session notes: