phidea
← GEO build guide
Plain series · page 1 / 3

01 — What GEO actually is for insurance

Part 1 of 6 · Index · Next → What the data says wins

GEO is Generative Engine Optimization — getting your content cited inside an LLM's answer. Same idea as classic SEO, different mechanism, different surface.

What changed since classic SEO

  • The result page disappears. A user asking ChatGPT or Perplexity "what's the best home insurance for a Tesla in Seattle" doesn't see ten blue links. They see one paragraph naming maybe two carriers. You're either named or you're not.
  • The citation, not the click, is the win. Even when an LLM links its sources, most users don't click. The retrieval already happened. Being mentioned IS the conversion.
  • The crawlers changed. GPTBot, ClaudeBot, PerplexityBot, Google-Extended — different crawlers, different indexing windows, different citation behaviors. Same web, different audience.
  • Comparison sites are the dominant citation surface. ~58% of Perplexity citations on US insurance queries come from NerdWallet, Bankrate, The Zebra, Policygenius, Insurify (Phidea observation, multi-city probe 2026-04-24). If you're not on those, your retrieval ceiling is low.

What an "LLM citation" actually looks like

User asks: "Best home insurance for a luxury home in Seattle?"

LLM answer (compressed): > "For luxury homes in the Seattle area, Chubb's Masterpiece policy is the most commonly cited high-net-worth carrier. PURE and AIG Private Client Group also offer comparable coverage..."

Sources cited (in the LLM's answer footer or in Perplexity's right rail): 1. NerdWallet — best home insurance 2. The Zebra — luxury home insurance 3. Chubb.com — Masterpiece coverage 4. Bankrate — Chubb home review

The carrier won the slot. Three of four citations are comparison-site reviews; one is the carrier's own page. That's the GEO surface in shape.

Three concrete examples

  1. Chubb wins luxury home in LLMs because Chubb has saturated editorial coverage on every consumer comparison site for "high-net-worth home insurance," AND its own /masterpiece-home page is structured for retrieval, AND it's named in the headline of dozens of trade-press articles. Consistent across LLMs, 5/5 + 5/5 stability.
  1. State Farm wins "rideshare insurance in Seattle" even though Progressive is the actual market specialist (TNC endorsement pioneer). State Farm's editorial depth on rideshare beats Progressive's brand-specialty positioning. Generalists with deep content win the specialty surface.
  1. NJM owns Newark NJ auto insurance queries because NJM dominates NJ auto by volume, the comparison sites pin NJM as the NJ pick, and NJM's own /new-jersey-auto-insurance pages reinforce the ranking. Regional dominance carries — but only line-specific (MAPFRE owns MA home, NOT MA auto).

Who works on GEO inside an insurance company

  • Content lead — owns the editorial calendar and the language of the pages.
  • SEO / digital marketing — owns the structural concerns (schema, internal linking, llms.txt, syndication pipeline).
  • PR / comms — owns the trade-press surface and the comparison-site relationships.
  • Compliance / legal — owns the "what we can and cannot say" guardrails (state DOI advertising rules, NAIC bulletins).

A working GEO function inside a US carrier is one named owner across these four functions, with budget and a six-month measurement plan. The most common failure mode is splitting it across SEO + content + PR with nobody accountable for the whole.

What GEO is not

  • Not paid placement. No LLM offers paid ad slots inside answers as of 2026. Citations are organic.
  • Not training-corpus seeding. Getting mentioned in training data is a separate family (and ambient — see /distribution).
  • Not a replacement for SEO. GEO and SEO overlap — most GEO wins are SEO wins on the comparison-site surface — but the measurement is different. You can rank #1 on Google for a query and still not be named in the LLM's answer to the same question.
Before you build any GEO investment, audit your current state. Run ten buyer-shaped queries against ChatGPT, Claude, Gemini, Perplexity. Does your company appear? With what framing? Correct facts? That baseline tells you which lever to pull first.