--- title: "How can health insurers win members when shoppers ask AI?" description: "By making plan facts, networks and ratings accurate and public before enrollment opens, because shoppers now ask AI tools to explain and compare coverage." canonical: "https://underneath.agency/resources/health-insurers-members-ai-search" published: 2026-10-07 updated: 2026-10-08 publisher: "Underneath (https://underneath.agency/agent)" entity: "https://underneath.agency/.well-known/entity.json" --- Guide · AI search # How can a health insurer win new members when shoppers ask AI to explain their plan choices? By making sure AI answers find accurate, current and consistent facts about your plans, networks and ratings before each enrollment window opens. Plan shoppers are confused, rarely compare, and increasingly ask chatbots about health and coverage. No study yet links AI answers to enrollment numbers, so treat this as a strong bet, not a measured channel. Nothing here is advice on choosing a plan. ## The short version 1. The individual market is large and partly new each year: 23.0 million people selected 2026 Marketplace plans, including 3.4 million new consumers, according to [CMS](https://www.cms.gov/newsroom/fact-sheets/marketplace-2026-open-enrollment-period-report-national-snapshot-2). 2. Medicare buyers face many options and mostly do not shop: the average beneficiary could choose among 42 plans in 2025, [KFF](https://www.kff.org/medicare/issue-brief/medicare-advantage-2025-spotlight-a-first-look-at-plan-offerings/) reports, and in an earlier KFF analysis 69% did not compare their coverage with other options during open enrollment. 3. Shoppers want help: in an [eHealth survey](https://s204.q4cdn.com/837903328/files/doc_news/Survey-75-of-Medicare-Beneficiaries-Say-Selecting-a-Plan-Is-Confusing-2025.pdf), 75% of Medicare beneficiaries called choosing a plan confusing, and an EBRI survey reported by [Health Populi](https://www.healthpopuli.com/2026/03/24/consumer-adoption-of-ai-for-health-and-self-care-doubling-to-36-in-a-year-via-rock-healths-latest-snapshot) found 42% of privately insured adults want AI tools to help them choose a health plan but do not know where to start. 4. OpenAI documents the use case: [ChatGPT Health](https://openai.com/index/introducing-chatgpt-health/) is described as helping people “understand the tradeoffs of different insurance options,” and OpenAI says over 230 million people a week ask ChatGPT health and wellness questions. 5. Google’s AI answers are hard to avoid in this category: in [our study of 1,248 US searches](https://underneath.agency/research/ai-overviews-frequency-study), financial services and insurance keywords triggered an AI Overview 88.0% of the time. ## Who buys health insurance, and what is a new member worth to an insurer? Three different buyers: individuals, Medicare beneficiaries and employers. Each won member tends to stay for years. **Individual market.** CMS counts 23.0 million 2026 plan selections on the Marketplaces. Of those, 19.6 million had 2025 coverage and selected a plan or were automatically re-enrolled, so the fight for new members centers on the 3.4 million newcomers and on returning members who actively shop. **Medicare.** Medicare Advantage enrolled 34.1 million of about 62.8 million eligible beneficiaries in 2025, or 54%, according to KFF data [summarized by TechTarget](https://www.techtarget.com/healthcarepayers/news/366628088/Top-stats-on-Medicare-Advantage-enrollment-costs-in-2025). The market is concentrated: UnitedHealth held 29% of enrollees and Humana 17%. Growth slowed to 1.3 million beneficiaries, or 4%, in 2025, so plans increasingly win members from each other rather than from new eligibles. **Employers.** The [KFF 2025 Employer Health Benefits Survey](https://www.kff.org/health-costs/2025-employer-health-benefits-survey/) puts the average annual premium at $9,325 for single coverage and $26,993 for family coverage. Employers usually choose through brokers and consultants, and their reasons for switching include their workers’ experience: in the [J.D. Power 2025 commercial member study](https://chaindrugreview.com/j-d-power-gap-widens-between-highest-and-lowest-performing-employer-sponsored-health-plans), 20% of employers cited low employee satisfaction as a top reason for switching health plans. Health startups selling to employers and plans face a related test, covered in [how healthcare startups compete with established brands](https://underneath.agency/resources/healthcare-startups-demand-ai-search). What makes a member valuable is how rarely people move. KFF found that 82% of Medicare Advantage drug plan enrollees did not compare their plan’s drug coverage with other plans in their area. The flip side: a shopper who leaves you during an enrollment window may be gone for many years. Carriers selling auto, home and other lines can see [how insurance carriers win quotes from AI](https://underneath.agency/resources/insurance-companies-customers-ai-search). ## Are plan shoppers already asking AI assistants about coverage? Yes for health questions in general; for plan choice specifically, the evidence shows interest more than measured use. The general shift is well documented. A [KFF poll](https://www.kff.org/health-information-and-trust/poll-finding/kff-tracking-poll-on-health-information-and-trust-use-of-ai-for-health-information-and-advice/) found 32% of adults turned to AI tools for health information and advice in the past year. Rock Health’s consumer survey, as reported by Health Populi, found use of AI chatbots for health information doubled from 16% to 32% between 2024 and 2025; ChatGPT was used by 23% of seekers and Gemini by 15%. On insurance itself, three signals point the same way: - **The platform documents it.** OpenAI lists insurance among ChatGPT Health’s uses, and one of its example prompts is “Based on my medical history, which of these insurance plans might be best for me?” OpenAI updated the page in July 2026 to say Health in ChatGPT is launching to US users 18 and older. - **Shoppers say they want it.** EBRI’s 2025 survey of 2,001 privately insured Americans aged 21 to 64 is the source of the 42% who want AI help choosing a plan. - **Brokers are building for it.** eHealth found 50% of Medicare beneficiaries would be interested in working with an AI agent by phone if it made shopping more efficient. eHealth sells plans and has an interest in that finding. None of these surveys measures how many enrollees asked ChatGPT or Gemini to compare carriers before enrolling. That figure does not exist yet. ## What do plan shoppers ask AI assistants? Mostly questions about plan types, networks, drugs, costs and ratings. We wrote these examples ourselves; none were collected from real shoppers. | Shopping concern | Illustrative question | |---|---| | Plan type | “What is the difference between an HMO and a PPO on the Marketplace?” | | Medicare choice | “What should I know when comparing Medicare Advantage and Medigap?” | | Network | “Which 2026 plans in Maricopa County include the Banner Health network?” | | Drugs | “How do Part D plans cover insulin in 2026?” | | Ratings | “Which Medicare Advantage plans near Tampa have high star ratings?” | | Extras | “Which plans in my area include dental and vision benefits?” | | Employer | “What should a 50-person company compare when switching group health carriers?” | Behind each question there may be several searches. Google says AI Overviews and AI Mode [may use a “query fan-out” technique](https://developers.google.com/search/docs/appearance/ai-features), issuing related searches across subtopics, and OpenAI says [ChatGPT search rewrites a question](https://help.openai.com/en/articles/9237897-chatgpt-search) into one or more targeted queries. A question about a county’s plans could become separate searches for networks, star ratings and drug coverage. When [our hidden-searches study](https://underneath.agency/research/ai-hidden-searches-study) logged ChatGPT’s searches, it looked for reviews or ratings in 46.2% of answers. A caution for insurers: AI answers on these topics are general information, not plan advice, and an assistant can get plan details wrong. That is why the accuracy of your public plan information matters so much. ## How does an AI answer turn into an enrolled member? By shaping which carriers and plan types a shopper checks before the enrollment window closes. As we understand the evidence, the path looks like this: 1. Before or during an enrollment window, a shopper or family member asks an assistant to explain options, check a doctor or drug, or compare carriers. 2. The answer names plan types, sometimes carriers, and cites sources such as government pages, insurer pages and comparison sites. 3. The shopper checks the insurer’s website, the official plan finder or a broker. In [Media Logic’s 2025 survey](https://www.medialogic.com/blog/healthcare-marketing/2025-medicare-aep-shopping-experience/) of 450 Medicare enrollees, insurer websites were the primary resource and brokers facilitated 58% of switches. 4. The shopper enrolls before the deadline; Medicare’s annual window runs October 15 to December 7, and HealthCare.gov’s 2026 window ran through January 15. 5. Retention follows. Because most people do not compare again, a member won this year is likely to renew. The windows make timing unusual. A software vendor can build visibility at any time; an insurer needs its plan facts to be right in AI answers in the weeks before and during enrollment. Media Logic found 20% of shoppers switched plans, and 64% of switchers changed insurers entirely. Those are the members up for grabs. Measurement is hard. A member who first heard your name in an AI answer may enroll through a broker or call center with no trace of that answer. We cover attribution in [how to prove GEO caused sales](https://underneath.agency/resources/prove-geo-caused-sales). ## What decides whether an AI assistant mentions your plans accurately? The platforms keep selection rules private; studies favor authoritative, consistent sources, and stale pages breed wrong plan facts. **Documented by the platforms.** Google and OpenAI confirm their AI answers search the web and link to sources. Neither explains how a carrier or plan is chosen. **Observed in studies.** - In ChatGPT’s answers to consumer health questions, [Jacques and colleagues](https://arxiv.org/abs/2601.17109) found 75.7% of 615 cited sources came from institutions such as medical centers, government agencies and Wikipedia. Commercial sites that were cited showed visible trust signals; our summary is in [what commercial health sites cited by ChatGPT have in common](https://underneath.agency/resources/chatgpt-cited-commercial-health-sites). - [Our pricing study](https://underneath.agency/research/ai-pricing-accuracy-study) found 61.9% of plan prices quoted by four assistants for software products were fully faithful to the official pricing page. Of 64 differing prices, 39 appeared on another page of the vendor’s own site. That study covered software, not insurance, but the lesson transfers, we infer: last year’s plan pages and outdated benefit summaries are likely sources of wrong answers. **What shoppers trust.** KFF found 77% of the public is concerned about the privacy of personal medical information given to AI tools. J.D. Power found members who understand their out-of-pocket and out-of-network coverage report higher satisfaction. Clear, accurate public explanations serve both the shopper and the assistant. **Our inference for insurers.** The facts shoppers verify (networks, formularies, costs, star ratings, extras) are mostly public, but they are often spread across PDFs, county-specific pages and old plan years. A benefit change buried in a PDF the assistant never reaches will not make it into the answer. A carrier whose current facts are easy to find, consistently named and supported by third-party ratings gives the assistant less room for error. No study has yet tested this with health plans. ## What does it cost an insurer to be missing or misdescribed in AI answers? Mostly lost consideration among the minority who actively shop, plus misinformation that sales teams must correct. - **The shoppers who matter are few and decisive.** If 69% of Medicare beneficiaries do not compare, the members you can win sit in the remaining 31%. eHealth found 51% intended to review coverage for 2026, down from 63% who said they did so the year before. - **Confusion is the norm.** In eHealth’s survey, 33% of beneficiaries said they did not have a good understanding of how Medicare Advantage, Medicare Supplement and Part D plans differ, and 33% wrongly believed Medicare covers GLP-1 drugs for weight loss. An AI answer that repeats a wrong benefit claim about your plan can cost a sale or create a complaint, we infer. - **Brand gaps are visible.** J.D. Power found regional satisfaction scores ranging from 594 to 523 on a 1,000-point scale. Ratings like these are public and can be cited, for good or ill. - **Answers shift.** The same question can return different carriers on different days; see [why AI answers about your brand change](https://underneath.agency/resources/why-ai-answers-about-your-brand-change). ## What GEO work fits a regulated health insurer? Accurate, current, consistent public plan information, backed by independent ratings, and reviewed by compliance. 1. **One clear identity per plan.** Use the same plan names, plan years and service areas on your site, in directories and in partner listings. When assistants mix up plan names or years, [fixing wrong brand information in AI answers](https://underneath.agency/resources/fix-wrong-brand-information-in-ai-answers) sets out the steps. 2. **Plan-year pages that can be read.** Publish benefits, networks and drug coverage as current web pages, not only PDFs, and mark or retire last year’s pages so they stop competing with this year’s. 3. **Plain explanations.** Write neutral explainers on plan types, enrollment windows and how to check a doctor or drug. Keep them educational; no personalized recommendations, and every claim checked against your filed plan documents. 4. **Independent proof.** Star ratings, accreditation and J.D. Power results are third-party sources assistants can cite. Report them accurately and with their year. The broader method is in [how brands build authority for AI search](https://underneath.agency/resources/how-brands-build-authority-for-ai-search). 5. **Reputation.** Reviews, complaint resolution and coverage in local news feed what assistants say about a carrier. 6. **Timed tracking.** Ask shoppers’ questions across ChatGPT, Gemini, Perplexity, Copilot, Claude and Google’s AI features before and during each enrollment window, more than once; [how many prompts to track](https://underneath.agency/resources/how-many-prompts-to-track-ai-visibility) explains the sample size. Medicare and Marketplace marketing are regulated, so treat AI-facing content as marketing material subject to your usual review. Do not seed misleading content or fake reviews; see [legitimate GEO versus manipulation](https://underneath.agency/resources/legitimate-geo-vs-manipulation). ## Which questions about AI and plan shopping are still unanswered? Whether AI answers change enrollment choices is unmeasured; most data show interest in AI help, not its effect. - **No enrollment data.** No survey counts members who used an AI assistant before choosing a plan. - **Some figures are older or partial.** KFF’s 69% comes from the 2021 open enrollment period. The eHealth and Media Logic surveys come from firms that sell plans or insurer marketing. - **Health citation research covers medical questions.** Coverage questions may draw on different sources than symptom questions. - **The software pricing study is not about insurance.** We use it as a pattern, not as a measure of insurer accuracy. - **Revenue links are thin in every industry, not only insurance.** The evidence that does exist is gathered in [does AI visibility drive business results](https://underneath.agency/resources/does-ai-visibility-drive-business-results). ## Where should a health insurer start before the next enrollment window? Check what the main assistants say about your plans in your top markets, then fix the public facts behind errors. Pick your largest counties and products. Write the questions an individual shopper, a Medicare beneficiary, an adult child helping a parent, and an employer’s broker would ask. Ask each one in several assistants, then ask again a few days later. Record which carriers are named, which sources are cited, and whether your networks, drug coverage, ratings and costs come back right for the correct plan year. Most errors in this industry trace back to outdated plan documents or inconsistent plan names. If you would like a second pair of eyes, [request an enrollment-season AI visibility review](https://underneath.agency/contact). It shows how AI answers describe your plans in the markets that matter most, where they cite outdated or wrong information, and which fixes are most likely to protect new member enrollment and renewals before the window opens. Our [generative engine optimization service](https://underneath.agency/services/generative-engine-optimization) maps that work onto a regulated insurer’s calendar, with readable plan-year pages, compliance review and tracking timed around each enrollment window. ## Frequently asked questions ### Do Medicare beneficiaries use ChatGPT to pick plans? No study measures it directly. eHealth found 50% of beneficiaries would be interested in an AI agent by phone, and OpenAI lists insurance comparisons as a ChatGPT Health use. ### Can an AI assistant recommend our plan to a specific person? Assistants can describe plans, but they are not licensed agents. Insurers should keep their own content educational and avoid personalized recommendations. ### Why do AI answers show last year’s benefits? Often because old plan-year pages remain online. In our software pricing study, 39 of 64 differing prices appeared on another page of the vendor’s own site. ### Do star ratings and J.D. Power awards help? They are independent sources assistants can cite, and ChatGPT searched for reviews or ratings in 46.2% of answers in our study. Their effect on mentions is untested. ### When should an insurer check its AI visibility? Before each enrollment window, and again during it. Medicare’s annual window runs October 15 to December 7. ## Sources - Centers for Medicare & Medicaid Services (2026-01-28), [Marketplace 2026 Open Enrollment Period Report: National Snapshot](https://www.cms.gov/newsroom/fact-sheets/marketplace-2026-open-enrollment-period-report-national-snapshot-2) - KFF (2024-11), [Medicare Advantage 2025 Spotlight: A First Look at Plan Offerings](https://www.kff.org/medicare/issue-brief/medicare-advantage-2025-spotlight-a-first-look-at-plan-offerings/) - KFF (2024-09-26), [Nearly 7 in 10 Medicare Beneficiaries Did Not Compare Plans During Medicare’s Open Enrollment Period](https://www.kff.org/medicare/issue-brief/nearly-7-in-10-medicare-beneficiaries-did-not-compare-plans-during-medicares-open-enrollment-period) - KFF (2025), [2025 Employer Health Benefits Survey](https://www.kff.org/health-costs/2025-employer-health-benefits-survey/) - KFF (2026), [KFF Tracking Poll on Health Information and Trust: Use of AI for Health Information and Advice](https://www.kff.org/health-information-and-trust/poll-finding/kff-tracking-poll-on-health-information-and-trust-use-of-ai-for-health-information-and-advice/) - TechTarget (2025), [Top stats on Medicare Advantage enrollment, costs in 2025](https://www.techtarget.com/healthcarepayers/news/366628088/Top-stats-on-Medicare-Advantage-enrollment-costs-in-2025) - Chain Drug Review (2025-05-28), [J.D. Power: Gap widens between highest- and lowest-performing employer-sponsored health plans](https://chaindrugreview.com/j-d-power-gap-widens-between-highest-and-lowest-performing-employer-sponsored-health-plans) - eHealth (2025-10-01), [Survey: 75% of Medicare Beneficiaries Say Selecting a Plan Is Confusing](https://s204.q4cdn.com/837903328/files/doc_news/Survey-75-of-Medicare-Beneficiaries-Say-Selecting-a-Plan-Is-Confusing-2025.pdf) - Media Logic (2025), [What We Learned from the 2025 Medicare AEP Shopping Experience Survey](https://www.medialogic.com/blog/healthcare-marketing/2025-medicare-aep-shopping-experience/) - Health Populi (2026-03-24), [Consumer adoption of AI for health and self-care doubling, via Rock Health’s latest snapshot](https://www.healthpopuli.com/2026/03/24/consumer-adoption-of-ai-for-health-and-self-care-doubling-to-36-in-a-year-via-rock-healths-latest-snapshot) - OpenAI (2026-01-07, updated 2026-07-23), [Introducing ChatGPT Health](https://openai.com/index/introducing-chatgpt-health/) - Google Search Central (2025), [AI features and your website](https://developers.google.com/search/docs/appearance/ai-features) - OpenAI Help Center (2025), [ChatGPT search](https://help.openai.com/en/articles/9237897-chatgpt-search) - Jacques and colleagues (2026), [Authority Signals in AI Cited Health Sources: A Framework for Evaluating Source Credibility in ChatGPT Responses](https://arxiv.org/abs/2601.17109), arXiv:2601.17109. - Underneath (2026), [When does Google show an AI Overview? 1,248 US searches](https://underneath.agency/research/ai-overviews-frequency-study) - Underneath (2026), [The hidden searches AI assistants run before they answer](https://underneath.agency/research/ai-hidden-searches-study) - Underneath (2026), [How faithfully do AI assistants quote software prices?](https://underneath.agency/research/ai-pricing-accuracy-study) --- This is the Markdown twin of https://underneath.agency/resources/health-insurers-members-ai-search. The HTML page is canonical. Publisher: Underneath, https://underneath.agency/agent. Site index: https://underneath.agency/llms.txt.