Answer Engine Optimization for Medicare & Health Insurance Agencies

What answer engine optimization means for a Medicare or health agency

When someone asks ChatGPT, Perplexity, or Google's AI Overviews a question about Medicare, the tool writes a paragraph and names a few websites it drew from. Answer engine optimization — AEO — is the work of making your website one of those named sources. It is not a new marketing channel you buy; it is the same website, rewritten so a machine can find the answer in it and quote it without ambiguity.

The mechanics are public, and unglamorous. Google's own documentation says there are no special requirements or markup for AI Overviews and AI Mode: a page has to be indexed and eligible to appear as a snippet, and standard SEO best practices are what qualify it³. OpenAI says allowing OAI-SearchBot is what lets a site surface in ChatGPT's search features⁴, and Perplexity says the same of PerplexityBot⁵. Past those qualifying conditions, no engine publishes a ranking formula — so the honest version of this work is: be retrievable, be answer-shaped, be obviously sourced, and be the kind of page an engine loses nothing by quoting. We wrote up the general shape of that in how AI engines decide which sites to cite and the SEO-versus-AEO distinction in the difference between SEO and AEO.

The questions Medicare and health shoppers actually ask an AI

Every vertical has a question shape, and Medicare's is unusually rigid: most of what people ask is about eligibility, timing, and mechanics, and most of it has a factual answer that does not depend on the agency answering it. That is the opening. These are questions an agency can answer completely and correctly, which is exactly the kind of page an engine can quote without hedging.

Notice what is missing from that list: "best Medicare plan," "cheap health insurance," and the rest of the commercial head terms agencies have chased for a decade. AI assistants rarely resolve those into a single cited source, because there is no defensible single answer. The citable surface is the boring middle — the questions with answers.

Why this vertical is different: the answer has to be compliant, not just correct

In every other insurance line, a page that answers a question well is simply a good page. In Medicare and health, the same page sits inside three overlapping rule sets at once: HIPAA if your agency handles protected health information, CMS's marketing rules if you or your marketing partners touch Medicare Advantage or Part D, and the FTC Act plus the state health-privacy and wiretapping statutes that apply whether or not HIPAA ever attaches.

That does not make AEO harder to do. It makes it harder to do carelessly, in two specific ways. First, the content: a page written to win a comparison question can slide into a claim about plans or carriers that your marketing review would never clear, and nothing in the publishing process stops it. Second — and this is the one that actually bites — the pages best suited to AI citation are the same pages where visitors type health information into a form. "Do I qualify for a Special Enrollment Period?" is a citable question and a data-collection event on the same URL.

The relevant rule is not exotic. HHS OCR's tracking-technology guidance says that a vendor receiving protected health information on a regulated entity's behalf is a business associate that needs a signed Business Associate Agreement, and that pages where visitors submit health information are where that trigger fires⁷. A federal court vacated one piece of that guidance in June 2024 — the automatic rule treating an IP address plus a visit to a public health page as PHI — and we walk exactly what survived in what OCR's tracking guidance actually says. Meanwhile Google's own policy states it does not offer BAAs for Google Analytics and tells HIPAA-regulated customers to use it only on pages that are not HIPAA-covered⁸, and Meta offers none for the Pixel. If a vendor will sign a confidentiality agreement but not a BAA, that is not a fix — see the difference between an NDA and a BAA.

The tracking problem sitting underneath the content problem

We stopped guessing about how common this is and measured it. On August 17, 2026 we scanned 1,286 U.S. insurance agency websites for client-side tracking technologies and published the aggregate dataset CC0¹. Half the sample — 647 sites — paired health-topic content with a lead form, which is the configuration regulators and plaintiffs' firms have spent three years litigating. Of those 647:

• 316 sites (48.8%) run at least one client-side tracking technology on those pages.
• 57 sites (8.8%) run a Meta Pixel — the tag with no BAA available at any price.
• 24 sites (3.7%) run session-replay tools, the technology at the center of current wiretapping claims.

Every one of those numbers is more likely an undercount than an overcount: 19.8% of the full sample runs a Google Tag Manager container whose contents the scan could not resolve, and we read served HTML on two pages per site. The full method and limits are in the study itself.

Here is why it belongs in an article about AI citation rather than a separate compliance post. AEO works by making your pages more findable, more quotable, and more visited. If the pages doing that work are leaking health-adjacent signals to ad platforms, the program's success and its exposure scale together. The fix order is therefore not negotiable: audit the tags on health pages first, then build the content. Our answer pages on which analytics tools an agency can legally use and whether consent banners actually fix it cover the stack decisions; the free Audit runs the same scan this study used against your own domain.

The speed floor every citation depends on

A page that cannot be fetched quickly and reliably cannot be quoted. On August 18, 2026 we ran Google's PageSpeed Insights API against 1,080 insurance agency homepages: the median mobile performance score was 61, the middle half of the industry sat between 52 and 70, and the median site took 7.9 seconds to paint its largest element against Google's 2.5-second "good" threshold for LCP². Only 12 of the 1,046 substantive sites — 1.15% — reached 99+, and 10 of those 12 were static builds.

Speed is not a ranking trick here; it is a precondition and a conversion issue. A Medicare shopper on a phone in a parking lot abandons a seven-second page, and a crawler on a budget gets less of your site per visit. What the 99+ tier has in common is architecture, not plugins — the detail is in can a real business website score 99+ on mobile and what score you actually need.

What a Medicare and health AEO build actually contains

Stripped of vocabulary, the build is six things, in this order:

1. A clean tag layer on health pages. No client-side ad pixels on quote forms or condition pages; measurement via a BAA-signing or self-hosted tool, or consent-gated server-side forwarding. This is first because it is the only item with legal consequences.
2. Answer pages on the questions above. One question per URL, the complete answer in the first sentence, a table or list with real values, a named and dated source for every number, and a visible "last reviewed" date. Our own answer library is built to that spec, so you can see the shape rather than take it on description.
3. County-level plan pages built from named plan-year data. This is the largest untapped surface in the vertical, because generic state pages answer nothing and most agency sites stop at a service-area list.
4. Machine-readable identity. Organization and Article markup with a real author entity, a licensed human named on the page, and FAQ markup that matches the visible text. Engines prefer sources they can attribute.
5. Agent access. An llms.txt that maps the site, AI crawlers unblocked at robots.txt and the firewall (which ones, and why), and where a calculator exists, a WebMCP endpoint an agent can call instead of scraping prose.
6. Measurement. Citation monitoring across the engines, because AI referrals mostly do not show up as ordinary search traffic — the mechanics are in why you can't see your AI search traffic.

What to do in the next 30 days

If you do nothing else: open your quote page in your browser's developer tools, filter the Network tab for facebook.com/tr and google-analytics.com, reload without touching the cookie banner, and watch what fires before consent. That ten-minute check tells you which side of the 48.8% you are on. Then write the five questions your office answers most often, one page each, answer-first, with sources.

Everything after that is volume and discipline. If you would rather see the whole picture scored first, the free Audit reads your site for AI-citation readiness and runs the tracking scan at the same time. If you sell more than one line, the sibling guides are here: AEO for life insurance agencies and AEO for commercial insurance agencies.

Questions Medicare and health agencies ask about AEO

Is answer engine optimization different from SEO for a Medicare agency?

The work overlaps heavily — Google's own AI-features documentation says standard SEO best practices are what make a page eligible to appear in AI Overviews, and OpenAI and Perplexity both say crawler access is the qualifying condition. What changes is the unit of success: SEO wins a blue-link position, AEO wins a sentence quoted inside someone's answer. For Medicare and health specifically, one more thing changes: the page that gets quoted has to be one you can defend, because the same page is a marketing asset and a compliance surface at the same time.

Can an AI engine cite a page that a compliance review would fail?

Yes, easily — engines evaluate retrievability and answer quality, not your carrier contracts or CMS marketing rules. That is exactly the risk. A page written to win a citation about plan comparison can carry a claim your upline would never approve, and unlike a paid ad, nothing forces it through review before it goes live. Treat AEO content as marketing material that happens to be machine-readable, and route it through the same review as anything else.

Do we have to remove Google Analytics to do this?

Not necessarily, but you have to decide where it sits. Google's own policy states it does not offer Business Associate Agreements for Google Analytics and instructs HIPAA-regulated customers to use it only on pages that are not HIPAA-covered. The practical pattern is: keep client-side ad and analytics tags off quote forms and health-condition pages, and use a BAA-signing or self-hosted tool, or consent-gated server-side forwarding, where you need measurement on those pages.

How long before AI engines start citing a new page?

Nobody publishes a number, and any agency that gives you one is guessing. What is documented is the sequence: the page has to be crawlable by the engine's fetcher, indexed and snippet-eligible for Google's AI features, and structured so the answer is extractable. In our own production work the first observable movement is usually in the user-triggered fetchers (ChatGPT-User, Perplexity-User) that pull a page live when someone asks, because those do not wait on an index refresh.

We are a small agency. Where does this actually start?

With eight to twelve answer pages on the questions your phone already asks you, each one answering the question completely in its first sentence, plus a tracking audit of the pages that collect health information. That combination is cheap, it is defensible, and it is the part most agency sites are missing — in our August 2026 scan, half the health-topic sites with a lead form were running client-side trackers on exactly those pages.