Using AI in insurance marketing

Your marketing is advertising, whoever writes it

The NAIC's model rule for life and annuity advertising (Model #570) defines an advertisement broadly: web pages, internet content, form letters, telemarketing scripts, sales aids and prepared sales talks and presentations, whether written by the insurer, its producers or third parties. An AI-drafted email, video, deck or call script used to sell falls inside it.

Each advertisement is the responsibility of the insurer and of the producer who created or presented it. Insurers must tell their agents at least once a year how to get approval for advertising the insurer didn't supply.

States have adopted different versions of this model or their own rules (NAIC state chart), and state unfair trade practices laws based on the NAIC model also ban misrepresentation and false advertising (Model #880). Your carriers' advertising guidelines tell you what applies to you.

What AI shouldn't write

• Anything untrue or misleading. Ads are judged by the overall impression they leave on an average person, not by the fine print.
• Anything that states or implies non-guaranteed elements are guaranteed. Show guaranteed and non-guaranteed values with equal prominence.
• Anything that makes life insurance or an annuity sound like something else. Call them "life insurance" and "annuity," don't use "investment," "savings plan" or "retirement plan" in a misleading way, and never "CD annuity."
• "Free" or "no cost," unless it is.
• Testimonials that aren't genuine, current and accurately quoted, or that hide a payment. Never publish AI-written reviews: the FTC's rule bans fake reviews and testimonials, including AI-generated ones (16 C.F.R. Part 465), and state insurance rules require testimonials to be real.
• Comparisons that are unfair or incomplete, statistics without a source, or ads that don't clearly name the insurer.
• Any mention of the state guaranty association as a reason to buy (Model #520).
• An AI-generated person in an ad without a clear disclosure, in New York since June 2026 (Cooley summary).

Telling people they're talking to AI

A handful of state laws require disclosing AI in some conversations. How they apply to an insurance agent's AI receptionist is largely untested:

Outbound AI voice calls also need consent under the TCPA. See Texting and calling clients.

The simplest approach works everywhere: have your AI assistant say it is an AI assistant at the start of calls and chats, and answer honestly if someone asks.

Annuity recommendations stay yours

The NAIC's best-interest standard for annuity sales (Model #275) has been adopted by every state except New York (NAIC, August 2025). It applies to recommendations to an individual consumer, not to general marketing or educational material.

Keep AI tools on the education and admin side. An AI that gives someone individual advice about buying or replacing an annuity is making a recommendation, with the disclosure form and documentation that come with it.

If you are a registered rep or adviser

• FINRA: "Firms are responsible for their communications, regardless of whether they are generated by a human or AI technology" (FINRA FAQ D.8). Chatbot output must be supervised like any other communication (Regulatory Notice 24-09).
• Retail communications need a principal's approval before use, and those about variable products are generally filed with FINRA within 10 business days of first use (Rule 2210). FINRA proposed a risk-based alternative in July 2026 (Notice 26-14); it is not in effect.
• Investment adviser reps: the SEC Marketing Rule covers testimonials and endorsements, and in 2024 the SEC fined two advisers for overstating their use of AI (SEC). Don't claim more AI than you use.

Insurers' AI rules can reach you

The NAIC's bulletin on insurers' use of AI (Model Bulletin) has been adopted in 25 states and DC (NAIC map, August 2026). It is addressed to insurers, but it covers marketing and distribution and expects oversight of third parties, so carriers may pass requirements to agents through their contracts.

Texas went further in June 2026 with a bulletin addressed to regulated entities and their agents: AI-supported decisions must follow insurance law, and consequential decisions need a person to review them (TDI Bulletin B-0003-26).

Keep records

• Insurers keep a copy of each advertisement for five years after its last use under the NAIC model; state periods vary.
• Producers keep the work papers and written communications for each policy sold, typically for the current year plus three, and some states require five (Model #910). Annuity recommendation records are kept too.
• Keep the final version of anything client-facing, the approval, and where and when it was used. For AI-made material, keeping the input you gave the tool helps if a question comes up later.

A checklist

• Treat everything AI produces as a first draft you are responsible for.
• Send client-facing material through your carrier's or IMO's approval process before it goes out.
• Check every number, rate and "guaranteed" against the carrier's own material.
• Never publish AI-written reviews or testimonials.
• Have AI assistants say they are AI, especially with callers in Maine and chat visitors in California.
• Keep AI on education and admin. Recommendations are yours.
• Keep copies of what went out and the approvals behind it.