Should your insurance business advertise in ChatGPT?

If you sell policies to people, yes. The questions are there in volume and a quote form converts like the sign-ups OpenAI publishes. If you recruit agents, run a $25-a-day test. If you sell to insurers, wait. Your buyers use paid plans that show no ads, and the two B2B tests with published numbers got zero leads.

Not every insurance business should be here. Some should be here today.

Spend on the wrong side of that line, and you pay $9 a click to reach interns on the free plan. Stay out on the wrong side, and a comparison site sits under every renters question in your state. Let’s talk about which side you are on.

The verdict, in one picture

The channel is built for a person with a question. The three kinds of insurance business have three kinds of buyer. So, three answers.

Selling policies: run it

The questions exist. In volume. And they are the questions that end in a quote.

On the largest public library, 6,075 insurance ads appeared across 6,163 researcher-written conversations. The question types: “best X for me” (20%), comparisons (15%), price (15%), long personal scenarios (15%), “fix this or switch” (15%), “is it worth it” (10%), short searches (10%). Every one of those is a buyer.

The conversion that follows is a form. Forms are what the platform has proven. OpenAI’s published customer results are sign-ups and leads: Jotform reports more than 10% of clicks became sign-ups at 60% lower cost than social and 30% lower than search. Angi reports cost per lead down 34% after switching to conversion bidding. LegalNature reports about 40% of visitors started a document and about 8% finished one.

Those are OpenAI’s numbers about its own customers. Treat them as the ceiling, not the forecast. But a renters quote is a lot closer to a Jotform sign-up than to a $40,000 software deal.

Here’s where I would start: one line of business, one state you are licensed in, two ads, the conversions objective billed on impressions. The setup is in Chapter 5.

Recruiting agents: test it

Producers do ask ChatGPT about carriers, commissions and appointments. I have not seen a public test of agent recruiting on the platform, so this one is judgment.

The honest way to hold it? A $25-a-day test against the number you already know. When we ran Rainbow’s go-to-market, Google Ads appointed agents at $88 each. That is the number a ChatGPT test has to beat or come close to. You will know in 14 days.

Two cautions. OpenAI tightened its rules on housing and employment ads in August, so an appointment offer should read as a business opportunity for a licensed professional, with the requirements stated. And the financial-services approval still applies. You are an insurer. Expect the licensure check.

Carriers and distributors: not yet

If your customer is an insurer, an MGA or a broker, the person who signs your contract runs ChatGPT on a Business or Enterprise plan. Those plans show no ads. You would be paying to reach the intern on the free plan.

The two B2B tests that published their numbers bear this out. Symphonic Digital spent about $500, got 53 clicks, 0 leads. Workshop Digital spent $2,319, got 739 clicks at $3.14 each, 0 qualified leads, while its Google Search campaigns produced 2 in the same period. Both were well-run tests by people who wanted the channel to work.

Spend the money on being the answer instead. When a VP of distribution asks ChatGPT which policy-admin vendors serve MGAs, you want to be named in the answer. No ad budget buys that. Our AI search work and the six steps to getting recommended are where a B2B dollar goes today.

The life moments nobody is buying

Brookings published the full dataset behind its audit: 3,602 ads, 191 advertisers, collected March 8 to April 12, 2026. I read all of it.

There is not one insurance advertiser in it. The study never asked about insurance, so that is not the surprise. The surprise is who showed up on the moments that create an insurance need.

A first car, a first apartment, a first house. That is where policies get bought. In spring 2026 the only brands in those conversations were selling parts and listings.

An insurer that describes those situations in its context hints, instead of the word insurance, is bidding on an empty lane. That matches what we saw in the insurance libraries, where a quarter of the prompts that triggered an insurer’s ad never said insurance.

What about health insurance and Medicare?

Approved US advertisers can run health insurance ads. OpenAI shows no ads in health-related conversations. Read those two sentences together.

The place a health plan wants to appear is the place the platform will not put it. A Medicare Advantage ad can only reach a 66-year-old asking about something else. Wait on health. Let the platform prove itself on a property or auto line first.

FAQs

We sell both policies and B2B software. Where do we start?

With the policies. Run the consumer test first and learn what a quote costs. Keep the B2B budget on organic AI visibility until OpenAI serves ads on paid plans.

Our agency recruits in 20 states. Can we test in all of them?

Test in one or two. A $25-a-day budget spread across 20 states learns nothing. Pick the states with the most open appointments and the cleanest tracking, and widen on day 15.

Does the 1.5x conversion claim from Criteo apply to insurance?

Criteo, OpenAI’s measurement partner, reported ChatGPT referrals converting about 1.5 times as well as other referral traffic for pilot retailers. Retail only, and a partner’s number. Use it as a reason to test, not a forecast.

What if our buyers are independent agents on the free plan?

Then they can see the ad, and the recruiting verdict applies: test it. Describe the appointment, the lines and the states in your hints, and count applications, not clicks.