Why does ChatGPT recommend quote sites, not me?

Because the national quote platforms are easier to verify and easier to quote than your website is. That's the whole answer. An assistant asked "who can help me with Medicare in Weber County" isn't running a popularity contest — it's assembling a response out of sources it can read, cross-check, and lift a clean sentence from. The big quote sites hand it all three. A typical agent site hands it a logo, a phone number and four paragraphs about personalized service.

That's a content and corroboration gap, not a size penalty — which matters, because gaps close and size doesn't.

What the quote sites actually have

Most agents assume it's ad budget. It isn't — none of the assistants sell placement in an answer. What the national sites have is coverage, consistency and corroboration:

• A page for every question, in every geography. Thousands of pages, generated at scale, each nominally about a plan type in a place. Thin, mostly — but they exist, and your county page probably doesn't.
• An unambiguous identity. Named the same way in directories, review platforms, filings and news coverage — so "is this a real company" gets the same answer from five independent places.
• Content in answer shape. Their explainers open by answering the question, because that's how snippets are won. Whatever quotes them reads the top of the page first and finds a complete sentence.

What Google and OpenAI actually say about this

Very little is public, and be suspicious of anyone claiming otherwise. Google's local ranking help names three factors — relevance, distance and prominence — and defines prominence as "how well-known a business is," based partly on "how many websites link to your business and how many reviews you have." That's the one documented line that genuinely favors a national brand, and it feeds the local results AI answers are often grounded in.

Google's guide to optimizing for generative AI features pulls in the other direction. It asks for "non-commodity content" and warns against material that "could easily be produced by a generative AI model" — which is a fair description of a mass-generated county page. It also describes query fan-out: the model issues "a set of concurrent, related queries" behind your one question. Those sub-queries are where a local agent gets a shot — several are narrower than anything a national page bothered to answer.

On the ChatGPT side the mechanics are simpler than the folklore. OpenAI's crawler documentation says OAI-SearchBot exists to "surface websites in search results in ChatGPT's search features," and that opted-out sites "will not be shown in ChatGPT search answers"; its help page on ChatGPT search says results are ranked on multiple factors and placement isn't guaranteed. Be crawlable, be useful, no guarantees. If your robots.txt blocks those bots, the rest of this page is moot — that decision is in should agents block GPTBot or ClaudeBot.

The gap, question by question

Here are the sub-questions an assistant resolves before it names anybody, and what each kind of site offers it.

That last row is the entire opportunity, and it's why we tell agents not to fight the national sites on national questions. You lose — and you don't need to win them. Nobody buys a plan off "best Medicare Advantage plan 2026."

Where a local agent already has the advantage

Ask a genuinely local question — which carriers have the most PPO options in your county, whether a specific clinic system is in-network for the plans sold there, who is licensed and takes appointments in a town of 12,000 — and the citable pool collapses. Often to nothing. The national page was written once for 3,000 counties. The other agents in your market never wrote anything down. Whoever publishes the real answer is the only source there is.

That's not a theory about ranking mechanics — it's what's left when you subtract everything a mass-produced page can't say. It's also perishable: the window is open because most agencies haven't moved yet.

Five things that move an agent into the answer

1. Say who you are in machine-readable terms. Organization and Service schema naming your agency, your license, your lines of business and your actual counties — not "the surrounding area." Which types belong on which page is in our schema map for agents.
2. Make your name, address, phone and service area identical everywhere. Site, Google Business Profile, carrier directories, association listings. Contradictions read as uncertainty, and uncertainty drops a source before ranking is a question.
3. Publish the county-level answers nobody else has. Real plan counts, carrier names and provider systems, sourced and dated — the content a national platform structurally cannot produce.
4. Rewrite your top pages answer-first. Heading is the question, first sentence is the answer, specifics instead of adjectives.
5. Get reviews that say something. Recent, specific, naming what you helped with. They feed prominence and they give a model words about you that you didn't write.

All five are doable alone if you have the hours; the long version is our AEO guide for insurance agents. It's also what clients pay us for, because every site we build ships with the schema, county pages and monthly articles already in this shape (what's included). Nobody can promise a citation. You control whether there's anything of yours worth citing.

One compliance note before you publish county pages

Those national platforms are third-party marketing organizations — and so are you, the moment you market Medicare Advantage or Part D plans for more than one carrier. Federal rules at 42 CFR 422.2267 require the TPMO disclaimer — "We do not offer every plan available in your area," the count of organizations and products you represent, and a pointer to Medicare.gov or 1-800-MEDICARE. More county pages means more pages that need it. Confirm exact wording and placement with your FMO and carriers; this isn't legal advice.

How to check where you stand

Ask ChatGPT and Perplexity the three questions a real prospect in your county would ask, and write down every source named — all national platforms is the normal baseline. Then narrow the questions to your county and carriers and watch the answers get vaguer. Vague is good news: it means the model has nothing local. Our 10-minute prompt test has the script.

Frequently asked questions

Why does ChatGPT recommend national quote sites instead of local agents?

Because those sites are easier to verify and easier to quote. A national quote platform publishes thousands of pages that answer plan questions directly, is mentioned across directories, review sites and news coverage, and states plainly what it does and where it operates — so an assistant handling a Medicare question finds a complete answer there that other sources corroborate. Most local agency sites offer a homepage, a phone number and a few paragraphs of adjectives, which gives an assistant nothing to lift and nothing to cross-check. It is a content-and-corroboration gap, not a penalty for being small.

Do AI assistants favor big brands over small local businesses?

Not automatically. Size helps with prominence — Google describes prominence in local ranking as how well-known a business is, based partly on how many websites link to it and how many reviews it has — but an assistant answering a local question needs locally specific, consistent facts, and a focused single-location business often has cleaner data than a national brand. Where small businesses lose is clarity: missing service areas, stale listings, vague descriptions of what they sell. Where they win is specificity a national page cannot match.

How do AI assistants decide which local business to recommend?

They look for a business they can verify from more than one place and describe accurately: consistent name, address, phone and service area across your site and the major directories; reviews that exist and are recent; plain statements of what you sell and where; and pages that answer the question asked in a form that can be quoted. Each platform runs its own retrieval and ranking pipeline and weights these differently, so the same agency can be named by one assistant and missing from another. None of them publishes a ranking formula, and none guarantees placement.

Do reviews affect whether an AI recommends my insurance agency?

Indirectly, yes. Google names review count as one of the inputs to prominence in local ranking, and assistants that ground their answers in web and map results inherit those signals. Reviews also give a model language about you that you did not write yourself, which is exactly the kind of corroboration it is looking for. A handful of recent, specific reviews naming what you actually helped with is worth more than a large pile of one-line five-star ratings.

Can a solo agent realistically beat a national quote site in AI answers?

For a national question such as which Medicare Advantage plan is best, no — and chasing it is a waste of money. For a question like who can help me compare Medicare plans in a named county, frequently yes, because the pool of citable local sources is thin or empty and the national page has nothing county-specific to say. Pick the questions where being local is the answer and answer them better than anyone else in your market. No one can promise a citation on any query.