AI search optimisation for healthcare: a practical guide

We sell AI search optimisation to clinics, so let's get the awkward bit out of the way first.

A lot of what agencies sell as "AI-readiness" simply isn't necessary, and Google says so itself. Its guide is blunt about the special files: "you don't need to create new machine readable files, AI text files, markup, or Markdown to appear in Google Search" (Google Search Central). It's just as blunt about chopping your pages into bite-sized bits: "There's no requirement to break your content into tiny pieces for AI to better understand it... There's no ideal page length" (Google Search Central).

So that AI-specific rewrite somebody quoted you for? Google isn't asking anybody for it.

The short version. Get yourself found in ordinary Google search first, because that's the identical work that gets you into the AI answers. They run off the same index. Then do the two clinic-specific things nobody else will mention. Check your `robots.txt` lets the AI search crawlers in, and check every review you publish on your own site against AHPRA's testimonial rule. That's the entire job.

The bit that's genuinely different for a clinic is narrower than the marketing suggests. It comes down to which "get more reviews" tactics you're actually permitted to use, and whether AI search visibility is even your real problem this quarter.

How a patient actually finds you now

In my own multidisciplinary clinic in New Zealand, everything came down to a single question. How easy were we to find and to book with?

A patient searched "osteo near me" or "physio near me", opened the map, and looked hardest at the closest few results. The Google reviews decided who got the click. From there it was book now, call, or a quick look at the website to see who the practitioners are and what's on the about page, then book or call.

Patients told me about the bad experiences they'd had elsewhere, and the examples were always small things. No online booking, for example, or a phone number that didn't work, or a booking button buried in the menu. Every one of those little problems cost somebody an appointment. Plenty of clinics still have no idea how to rank in Google in the first place.

Now a chunk of that search is starting somewhere else. Here's an example of how it goes. A patient describes a sore knee to ChatGPT, Gemini, Perplexity or Claude before anyone's named a diagnosis, then asks it for a physio nearby who treats that. It's a much longer conversation before they ever reach your site, so whatever they find when they land has to hold up.

Is it worth chasing? Here are two numbers, and both of them come with strings attached.

Semrush found the average visitor arriving from AI search converted 4.4 times better than one from ordinary search (Semrush, 21 July 2025). Semrush sells AI visibility software, so that's a vendor talking about its own marketplace. The study also covered digital marketing topics rather than healthcare, and the page doesn't fully show how they got to the number.

HubSpot says it saw 3 times better conversion from AI-sourced leads in 2025 (HubSpot). That figure comes from HubSpot's own software business, not from anybody's clinic. I've worked with HubSpot, so treat that as a disclosed relationship, not a neutral pick.

Treat both as a hint rather than a promise. My own guess is that a patient who's talked it through with AI first turns up better matched to what you actually treat. That's a guess, and I'll call it one.

What actually gets you shown

Before any of the AI business, your page has to clear the original basics, and there are only a couple of them.

An AI Overview is the summary Google writes at the top of the results page, with a few links tucked underneath it. To be one of those links, your page has to be "indexed and eligible to be shown in Google Search with a snippet" (Google Search Central). Indexed means Google already has your page saved in its library. A snippet is the grey line of text under your blue link. If you don't get that far in normal search, you're not going to turn up in the AI answer either.

The reason is that Google's AI features run "on our core Search ranking systems to retrieve relevant, up-to-date web pages from our Search index" (Google Search Central). Same engine underneath, different wrapper on top.

So the boring housekeeping still matters, and two particular items matter most.

A Google Business Profile is the free listing that puts you on the map with your hours, phone number and reviews. Yours needs to be accurate and genuinely complete. A practitioner in a public-facing role can have their own profile, titled with just their name, kept separate from the clinic's location profile. What you shouldn't do is run several profiles for one person's different specialities (Google).

An XML sitemap is a file listing every page on your website, so a search engine never has to guess. Keep it current and accurate. Bing makes the same point from its own side, saying that accurate sitemaps, 301 redirects and IndexNow all feed classic search and "AI-generated experiences, grounding results, and citations" (Bing).

Then there's the part that isn't mechanical at all, which is writing something only you could have written. Google draws a line between "commodity content" and writing that comes from real, first-hand experience. It says the second one "will likely influence your website's presence in generative AI search... more than any of the other suggestions in this guide" (Google Search Central). That's Google's own ranking of what matters, and it's sitting at the top.

Health pages are held to a higher standard again. Google's quality raters treat anything that can affect your health as "Your Money or Your Life", and apply "very high PQ [Page Quality] standards" to it. The worked example in their own guidelines is a page about the flu (Google Search Quality Rater Guidelines). The AI features lean on the same ranking systems, so the same bar follows you there.

Reviews are what convince them once they've found you

Getting discovered is only half of it, because reviews do everything after that. That hasn't changed and it isn't going to. But how you ask has rules attached, and they apply whether AI sent the patient or Google did.

You can't pay for reviews. Google bans incentives "such as payment, discounts, free goods and/or services" in exchange for a review, and it strips incentivised ones out of Maps. Just asking is fine, as long as you don't tell them what to write or lean on them while they're standing at the desk (Google). That's Google's own rule, not local law, so it reads the same in Australia and New Zealand.

When you ask matters as much as asking. In my clinic the sweet spot was after a patient's first or second session, once they'd had enough of the care to write about. We automated it. A message went out from the practice management system at that same point, for every patient, so nobody at the front desk had to remember.

Ask every patient, not just the ones who seemed pleased. Google also bans clinics from "selectively soliciting positive reviews" and from discouraging negative ones (Google). Choosing the right moment is perfectly fine, but choosing the right people definitely isn't.

The extra layer if you're a regulated health service

This is where everything gets specific to you, and where the generic marketing advice quietly gets you into trouble.

A testimonial, in AHPRA's sense, is a recommendation or positive statement about the clinical side of your service. That covers a symptom, a diagnosis, a treatment or an outcome, and none of it can go in your advertising. Comments about "customer service or communication style" aren't testimonials and you can use them (AHPRA).

You're also on the hook for anything you republish. You can put Google reviews on your own site, but whoever controls the site answers for them, and picking out only the glowing ones "has the potential to be misleading and breach the advertising requirements" (AHPRA).

None of this touches what a patient writes on Google off their own bat. It only bites once you lift that content into your own advertising.

New Zealand osteopathy is stricter, and we found no carve-out. The Osteopathic Council's advertising policy bans "testimonials whether from patients or any other person", full stop. Its 2023 Code of Conduct still points back to that policy (OCNZ). Reading that as covering embedded Google reviews on your own site is a fair call, not something the policy spells out word for word. We haven't checked the NZ physiotherapy equivalent, so don't assume it matches.

Letting the AI bots in, and whether that's your problem

A crawler is a bot that reads your website on behalf of a search engine or an AI tool. Your `robots.txt` is a plain text file at the root of your site that tells each bot what it's allowed to look at. One line per bot.

Now the bit worth knowing. OpenAI, Perplexity and Anthropic each split "can we cite you in an answer" from "can we train on you", so that's two separate bots with two separate switches.

Block the wrong one and you've quietly shut yourself out of the answer entirely. Consider this a five-minute check rather than an actual project.

But only bother if being found is genuinely your bottleneck. If your real problem is administration, or hiring people, `robots.txt` isn't going to fix anything. For instance, if your team is drowning in manual, repetitive paperwork, that's a different job entirely. AI workflow automation is the work for that, not search.

The part clinics still get wrong

Here's the mistake I see most often. A clinic does all the work to get discovered, then loses the patient anyway because the website looks dated or the formatting is a mess. Something doesn't quite line up, and the patient clicks away.

It's the same failure as the dead phone number from earlier, just later in the chain. All that ranking work gets undone by something as small as a broken menu.

A practical checklist

• Check `robots.txt` allows OAI-SearchBot, PerplexityBot and Claude-SearchBot, separately from GPTBot and ClaudeBot. Googlebot is implicit and covers AI Overviews.
• Search for your page in ordinary Google first. If it doesn't appear with a snippet, nothing else on this list matters yet.
• Keep an accurate XML sitemap with current `lastmod` dates.
• One Google Business Profile per practitioner where that applies, kept separate from the clinic's own.
• Make sure the online booking works, the phone number connects, and the booking button isn't hidden inside a menu.
• Look at your website the way a stranger would. Dated design and sloppy formatting cost you patients who have already found you.
• Automate the review ask at a fixed point, such as after the second session, and send it to every patient. No incentives, no filtering, no telling them what to say.
• Check every review you republish against AHPRA's clinical test, or OCNZ's blanket ban in New Zealand.
• Skip `llms.txt`, chunking and AI-specific rewrites. Google says none of it is needed.
• Write from what you've actually experienced yourself, because nobody else can copy that.

FAQ

What is AI search optimisation?

Making your site easy for an AI tool to read, understand and quote when someone asks it for a recommendation. It sits on top of the SEO that already gets you found in ordinary search. You'll also hear it called generative engine optimisation, or GEO. Same thing, fancier name.

How do I show up in ChatGPT?

The same fundamentals that get you found in Google. That means an indexed page that appears with a snippet, an accurate Google Business Profile, and content written from genuine experience. Then check `robots.txt` lets OAI-SearchBot in.

Is it different from SEO?

Mostly it isn't. Google's own guidance says the same groundwork carries over, and names `llms.txt`, chunking and AI-specific rewrites as things you don't need (Google Search Central).

Do I need an llms.txt file?

No. Google doesn't use one to decide what appears in Search or in its AI features, and it says so in plain words (Google Search Central).

Can I use patient reviews to prove my results to an AI?

Not in your own advertising, certainly not if the review touches clinical results. AHPRA bans testimonials that describe a symptom, diagnosis, treatment or outcome. Comments about service or communication style are perfectly fine to use. That restriction lands on you using the content in advertising, not on a patient posting their own review on Google. AHPRA's rule doesn't reach the patient themselves. New Zealand osteopathy is stricter again, with no non-clinical carve-out we could find (AHPRA, OCNZ).

Where to start, and where Opal picks it up

If you're not showing up in ordinary Google search yet, start there. AI visibility is built on the same foundation, so you get both of them out of the one piece of work.

If you are showing up already, go and read your own site like a patient who's never heard of you, then do the five-minute `robots.txt` check.

That's the work we do in AI search and growth for clinics. For where AI fits across the rest of the clinic, beyond the marketing side, see AI in healthcare administration.