AEO for Healthcare

Something changed in healthcare search over the past eighteen months, and most medical practices didn’t notice. By December 2025, 88% of healthcare queriestriggered a Google AI Overview — up from 59% just two years earlier. Treatment and procedure queries hit 100%. That means when a patient types “what are the side effects of metformin” or “how long does a root canal take,” Google now answers the question directly at the top of the page, and more than half of those searchers never scroll down to click anything.

Meanwhile, 40 million people ask ChatGPT a health question every single day. One in four of ChatGPT’s 800 million regular users submits a medical prompt weekly. Seven out of ten of those conversations happen outside clinic hours — at midnight, over the weekend, during a moment of anxiety when calling a doctor isn’t an option.

The shift from “Dr. Google” to “Dr. AI” isn’t coming. It already happened. And it introduces a problem that no other industry faces quite the same way: healthcare content lives under YMYL rules — Your Money, Your Life — which means AI engines hold it to the strictest citation standards on the internet.

Who actually gets cited in health AI answers?

Not who you’d expect. The top four domains cited in Google AI Overviews for health queries are Healthline (113,728 mentions), Cleveland Clinic (99,680), Mayo Clinic (89,103), and WebMD (87,025). These aren’t the largest hospitals or the most famous doctors. They’re the organizations that invested earliest in structured, clear, evidence-based patient education content — written for humans, formatted for machines.

That last part matters. Cleveland Clinic didn’t earn nearly 100,000 AI citations by writing brilliant prose. They earned them by doing something most healthcare organizations still haven’t done: they structured their content so AI engines could extract clean answers. Every condition page has a consistent format — definition, symptoms, causes, treatment, when to see a doctor — with proper headings, schema markup, and a “Frequently Asked Questions” section at the bottom. That FAQ section alone, marked up withFAQPage schema, accounts for a disproportionate share of their citations.

Perplexity, interestingly, handles health differently. In comparative studies across medical conditions, Perplexity consistently scored highest for quality and reliability because it surfaces peer-reviewed journal articles and current clinical protocols. If your practice publishes content that references specific studies with citations, Perplexity notices.

How YMYL became the AI gatekeeper

YMYL started as a Google quality rater guideline. It told human evaluators to hold health and financial content to a higher standard. But over the past two years, it evolved into something bigger: YMYL is now the gatekeeping standard that every AI platform applies to medical content before deciding whether to cite it.

The practical effect is that AI engines require three trust layersbefore they’ll cite healthcare content:

This is the core tension in healthcare AEO: the content that patients want (simple, reassuring, accessible) and the content that AI trusts (evidence-backed, professionally authored, institutionally verified) are often created by different teams with different goals. The organizations that win AI citations are the ones that bridge both.

What small practices consistently get wrong

After analyzing hundreds of healthcare websites — from solo dental practices to multi-location urgent care chains — the same five gaps appear over and over. They’re not exotic problems. They’re oversights that exist because most healthcare marketers optimized for Google’s blue links, not AI’s citation engine.

The “About Us” page is the only page with provider credentials

AI engines look for author attribution on the content itself, not on a separate bio page. If Dr. Sarah Chen wrote your guide to managing type 2 diabetes, her name, credentials, and a link to her profile need to appear on that specific page — ideally near the top, reinforced with Physician schema. Most practices bury provider information three clicks deep, where no AI crawler connects it to the content.

FAQ sections exist but lack schema markup

Plenty of healthcare sites have FAQ pages. Almost none mark them up withFAQPage structured data. The FAQ text is visible to humans who visit the page, but invisible to AI engines scanning for structured question-and-answer pairs. This is the single fastest fix in healthcare AEO — often implemented in under an hour — and it has an outsized effect because FAQ schema snippets achieve an 87% click-through rate when surfaced.

Medical content uses marketing language instead of clinical language

“Our state-of-the-art facility provides world-class orthopedic care” tells AI nothing. “We treat ACL tears, meniscus injuries, and rotator cuff tears using both arthroscopic and open surgical techniques” tells AI exactly which queries to match your page against. AI engines can’t assess quality claims (“world-class”), but they can map specific conditions and procedures to user queries with high confidence.

No MedicalEntity schema anywhere

Most healthcare websites have basic Organization or WebPageschema at best. They’re missing the entire medical schema vocabulary:MedicalClinic, Physician, MedicalCondition,MedicalProcedure. This vocabulary exists specifically so AI engines can understand healthcare entities with precision. Without it, your orthopedic practice looks the same as a shoe store to a machine.

Content hasn’t been updated in years

Healthcare AI citations have an unusually strict freshness requirement. 95% of ChatGPT citations come from content updated within the last 10 months. A treatment guide published in 2022 — even if still medically accurate — is unlikely to be cited because AI can’t verify whether guidelines have changed since publication. Adding a lastReviewed date and updating content quarterly is table stakes.

The schema stack that healthcare sites actually need

Healthcare schema is more complex than most industries because medical entities have their own dedicated vocabulary in schema.org. Here’s the practical hierarchy, ordered by implementation priority.

The common mistake is implementing these in reverse order — spending weeks on MedicalCondition schema for 50 conditions while ignoring the FAQ and Physician schema that would generate citations immediately. Start with FAQ and clinic/provider schema. Those cover the queries patients actually ask AI.

HIPAA and AEO: what you can and can’t optimize

The question comes up in every healthcare AEO conversation: “Does optimizing for AI violate HIPAA?” The short answer is no — but only if you understand the boundary.

AEO optimization involves making publicly availablehealthcare information more discoverable and citable. You’re structuring content about conditions, procedures, insurance plans, and provider credentials — none of which involves Protected Health Information (PHI). The line is clear:

If you use AI tools internally to help draft patient education content, that’s fine — just don’t paste patient records into ChatGPT. If you feature patient stories on your website, they must be fully consented and de-identified, with a signed release that covers AI indexing. The safest approach is to treat AEO optimization as a public communications exercise, not a clinical one.

How a small practice competes with Mayo Clinic

You can’t outpublish Cleveland Clinic. They have hundreds of medical writers producing thousands of condition pages. But you can outperform them on the queries that matter most to your practice: local health queries.

When someone asks AI “best dermatologist in Austin for eczema,” Mayo Clinic is irrelevant. The AI needs a local provider with dermatology credentials, patient reviews mentioning eczema treatment, and a website that confirms the specialty. This is where small practices have an asymmetric advantage — because almost none of them have implemented any AEO at all.

The first family dentist in a zip code to add Physician schema with board certification details, FAQPage schema addressing common patient fears, and a MedicalClinicschema with insurance and hours will capture a disproportionate share of local AI health recommendations. Not because they’re the best dentist. Because they’re the only one AI can confidently recommend.

A realistic four-week implementation plan

This is what a solo practice or small clinic can accomplish without hiring a developer or an agency. One focused task per week.

What types of healthcare content AI actually cites

Not all content is created equal in healthcare AEO. After analyzing citation patterns across ChatGPT, Perplexity, and Google AI Overviews, clear winners emerge — and they’re not the content types most practices prioritize.

Patient education guides that follow a consistent structure (what is it → symptoms → causes → treatment → when to see a doctor) get cited most often. This is the Cleveland Clinic formula, and it works because the structure mirrors how AI engines parse medical content. Each section maps cleanly to a different query type.

FAQ pages with specific, local answersperform disproportionately well for small practices. “Do you accept Blue Cross in Texas?” and “What’s the wait time for a new patient appointment at [practice name]?” are queries that Cleveland Clinic can’t answer for your patients. When these questions have schema-marked answers on your site, AI cites you by default.

Procedure explainers with honest recovery timelinesearn trust. Patients ask AI “how long is recovery from [procedure]?” constantly. Pages that give specific ranges (“most patients return to work in 5-7 days; full recovery takes 6-8 weeks”) with citations to clinical data get cited far more than vague promises (“quick recovery with our advanced techniques”).

What doesn’t get cited: practice news (“We won a Best Of award!”), generic wellness blog posts without clinical depth, and provider marketing that reads like ad copy. AI engines are remarkably good at distinguishing medical education from medical marketing, and they strongly prefer the former.

The emerging concept of an “AI trust score” for healthcare

No AI platform publishes a formal trust score for healthcare sites. But the citation patterns reveal a clear hierarchy that functions like one. Sites earn AI trust through a combination of signals that compound over time:

Consistent NAP (name, address, phone) across every platform. Reviews on Google, Healthgrades, and Zocdoc that mention specific conditions and providers by name. Schema markup that confirms specialties and credentials. Content that references clinical studies and carries visible authorship. A lastReviewed date within the past year.

None of these signals alone is sufficient. Together, they create a profile that AI engines treat as authoritative for specific medical queries in a specific geographic area. The practices that build this profile first in their local market will be very difficult to displace — because AI recommendations, unlike Google rankings, tend to be sticky. Once an AI learns to trust a source, it keeps citing it until a stronger signal contradicts it.