The State of AI Visibility for Australian Clinics 2026

Most Australian clinics look ready for AI search but are not. In a 2026 MaxConnex study, 94% of clinic homepages carried some structured data, yet only 56% carried the schema type that AI answer engines actually use, and directories dominated 64% of patient-style AI searches. The gap between looking ready and being ready is the story of this report.

This is original research. We ran two connected studies in 2026: one measuring who actually gets named when a patient asks an answer engine for a clinic recommendation, and one auditing the structured-data signals on real Australian clinic homepages. The findings are reported as aggregates only. No individual clinic is named, praised or criticised anywhere in this report.

Key findings

Why this matters

Answer engines now sit between patients and clinics. When someone asks ChatGPT, Perplexity or Google AI Overviews for a good GP, dentist, psychologist, physiotherapist or skin clinic in their city, the reply is not ten blue links. It is a short, confident shortlist of names. Whoever gets named in that shortlist wins the first, and often the only, consideration a patient gives.

Our first study shows that today those names are heavily skewed towards directories and aggregators, not clinic websites. In almost two thirds of the patient-style searches we ran, the top of the answer was owned by HealthEngine-style booking platforms, review aggregators, forum threads and editorial "top 10" listicles. Clinics were present, but they were rarely the primary source the answer engine leaned on.

Our second study explains part of why. A clinic that is structured for AI, with the correct schema type, clear entity data and directly answerable content, gives an answer engine everything it needs to name that clinic with confidence. A clinic that ships generic markup, or none of the signals that matter, forces the engine to fall back on the directories that do. The encouraging news for clinic owners is that the strongest signals are also the most neglected, which means the room to leapfrog is real and largely uncontested.

Part A: who wins when patients ask AI for a clinic

We ran 25 patient-style searches, five verticals across five capital cities: Sydney, Melbourne, Brisbane, Perth and Adelaide. The verticals were general practice, dental, psychology, physiotherapy, and cosmetic and skin. Each search used a natural "best {vertical} in {city}" style query with an Australian location, and we recorded the top eight results.

Every single query, 100% of them, surfaced at least one specific local clinic website in the top eight. So clinics are not invisible. The problem is prominence. In 64% of queries, three or more of the top five results were directories, aggregators, forums or editorial listicles rather than clinic websites. On average only 4.04 distinct clinic websites were nameable per query, with the remaining slots absorbed by non-clinic sources. The pattern varied sharply by vertical.

Searches dominated by directories, by vertical (out of 5)

The split is instructive. Verticals with mature booking platforms and association directories, such as general practice, dental and psychology, are the hardest for a clinic to own directly, because the aggregators are strong and well structured. Cosmetic and skin, where individual clinics invest more in their own websites and branded content, was the one vertical where clinic sites, not directories, tended to win. That is a preview of what happens across every vertical once clinics structure their own presence properly.

Part B: are clinics ready for AI search

For the second study we audited 50 real Australian clinic homepages, ten per vertical, and checked the server-rendered HTML for five signals that answer engines rely on. We looked for any JSON-LD structured data, a LocalBusiness or medical-specific schema type, FAQPage schema, a real llms.txt file, and a non-empty meta description.

AEO-readiness signals present, across 50 clinic homepages

We defined a homepage as "AEO-ready" only when it met three conditions together: it had JSON-LD, it had a LocalBusiness or medical-specific type, and it had a meta description. On that stricter test, 56% of clinics, 28 of 50, qualified. Readiness was uneven across verticals, and the leaders were not the ones you might expect.

Cosmetic, skin and physiotherapy clinics led on readiness at 70%, which lines up with Part A: the vertical that owned its own AI results was also the vertical investing most in its own web presence. General practice and psychology trailed at 40%, and those are exactly the verticals where directories dominated the AI answers. The correlation is not proof of causation, but the direction is consistent and hard to ignore.

The schema gap

This is the sharpest finding in the report. Almost every clinic ships some structured data, 94% of them, so on a surface audit they all look covered. But only 56% ship a LocalBusiness or medical-specific type. The rest ship generic markup, most commonly a bare Organisation or WebSite block, which tells an answer engine that a website exists but says almost nothing about what kind of clinic it is, where it operates, or what it treats.

Answer engines lean on the specific type to decide whether a business is the right answer to a local, medical query. A generic Organisation tag is close to invisible for that purpose. So the real readiness number is not the reassuring 94%, it is the 56%, and the difference between them is pure, recoverable upside.

FAQPage schema is the biggest untapped lever of all. Only 10% of clinics use it, yet it is the single most direct way to feed an answer engine ready-made question-and-answer pairs it can quote verbatim. A clinic that publishes clear FAQ schema is handing the AI the exact words to say about it. Ninety per cent of clinics are leaving that on the table.

What clinics can do

None of the fixes below require a rebuild, and most are one-off changes that keep paying off. In order of impact:

• Ship the correct schema type. Replace or upgrade generic Organisation markup with a LocalBusiness or medical-specific type that names your services, practitioners, location and hours. This is the change that moves you from the 94% who look ready to the 56% who are.
• Add FAQPage schema. Turn the questions patients actually ask into structured question-and-answer pairs on your key pages. With only 10% of clinics doing this, it is the fastest way to stand out to an answer engine.
• Keep Google Business Profile and directories consistent. Answer engines cross-check your name, address and phone across sources. Consistent, complete listings turn the directories that currently outrank you into signals that reinforce you.
• Publish a genuine llms.txt file. Give AI crawlers a clean, plain-language summary of who you are and what you offer. Only 28% of clinics have one, and many of those are auto-generated stubs rather than deliberate files.
• Write pages that answer questions directly. Lead with the question, give a clear 40 to 60 word answer, then expand. Directly answerable content is what an engine can lift into its reply.

The clinics that win AI visibility in 2026 are not the ones with the most content. They are the ones whose structured data tells an answer engine, unambiguously, exactly what they are and who they help.

Methodology

Part A measured AI recommendation visibility using 25 patient-style searches: five verticals, general practice, dental, psychology, physiotherapy, and cosmetic and skin, across five Australian capital cities, Sydney, Melbourne, Brisbane, Perth and Adelaide. Each search used a natural "best {vertical} in {city}" style query with an Australian location set through a web-search proxy, and we recorded the top eight results. A search was counted as directory-dominated when three or more of its top five results were directories, aggregators, forums or editorial listicles rather than individual clinic websites.

Part B audited 50 real Australian clinic homepages, ten per vertical. For each homepage we fetched the server-rendered HTML and checked five signals: any JSON-LD structured data, a LocalBusiness or medical-specific schema type, FAQPage schema, a real llms.txt file at the site root, and a non-empty meta description. A homepage was classed as AEO-ready when it had JSON-LD, a LocalBusiness or medical type, and a meta description together. Both studies were run in 2026 and report vertical and city level aggregates only.

Limitations

We want this research to be useful, so we are direct about what it can and cannot claim.

• The sample is purposive, not random. Clinics were drawn from those that already rank for their vertical and city, so the set skews towards established, marketing-active practices. True population readiness across all Australian clinics is very likely lower than the figures here.
• Web search was used as a proxy for AI answers. It is a reasonable stand-in for what answer engines surface, but it is not a literal capture of a ChatGPT or Perplexity response, which can vary by user, session and time.
• Schema detection was server-rendered-HTML only. Structured data injected later by JavaScript would be undercounted, so the schema figures are conservative and the real presence of some signals may be slightly higher.
• Many llms.txt files were auto-generated by SEO plugins rather than written deliberately. The 28% figure therefore overstates genuine, considered effort on that signal.
• "Directory-dominated" uses a stated threshold, three or more of the top five results being non-clinic sources. A different threshold would move the headline share up or down, so the 64% should be read together with its definition.

Read with those caveats in mind, the shape of the findings is clear and consistent: clinics are present in AI results but rarely prominent, directories currently absorb the visibility, and the structured-data signals that would change that are the ones most clinics have not yet shipped.