AI Search Shift
By the time a patient picks up the phone to book with a practice, they’ve often already decided. The research happened before the call. What they read, in what order, and whether your practice appeared at all in that sequence, is largely invisible to you unless you look for it deliberately.
The research trail a patient follows before choosing a physio has three layers. The first is the review stack that appears in local search and sets the short list. The second is the practice website content that either confirms or undermines the initial impression. The third, and increasingly significant for new patient acquisition, is AI summaries that synthesise available information into a named recommendation before the patient reaches a results page at all. Most practices are actively managing none of these three layers with any deliberateness.
The review stack that sets the short list
When a patient searches for a physio in their suburb, the practices that appear in the local pack are the ones with the strongest combination of relevance, proximity, and prominence. Reviews are the most controllable prominence signal. Practices with 80 or more Google reviews and a 4.7 or higher average dominate over practices with 15 reviews and a 4.5 average, even when the actual care quality is comparable. The quantity and recency of reviews both matter: a practice adding two or three new reviews per week outperforms a practice with 87 lifetime reviews and nothing added in the last six months.
The AHPRA compliance layer on reviews is specific and worth understanding clearly. Practitioners can encourage patients to leave reviews, but cannot direct them toward clinical content, selectively showcase positive reviews in advertising, or repurpose reviews that discuss clinical outcomes onto their own website. As covered in our piece on Google reviews and AHPRA, the rules leave significant room for review cultivation when approached correctly. The compliance risk is in how the reviews are used, not in asking for them.
What the practice website actually communicates
Once a patient has identified a practice from the review stack, they look at the website. The specific things they’re checking, usually within the first 60 seconds: whether the practice clearly handles their specific condition or activity, whether there’s a practitioner who sounds right for their situation, whether booking looks simple and immediate. Generic “quality physiotherapy” homepage copy fails every one of these tests. Specific condition pages, detailed practitioner profiles, and a booking button that’s visible without scrolling pass them.
The decision to contact a practice is often made or abandoned at the website, not at Google. A strong review profile that lands a patient on a generic, uncredentialled website loses the patient at the second step. The website’s job at this stage isn’t to persuade: it’s to confirm what the review stack suggested. Specificity, not polish, is what does that work.
A useful test: open your practice website as if you’ve never seen it before and have a specific injury, say a running-related knee problem. Within 60 seconds, can you confirm the practice handles that type of complaint, identify a practitioner who looks relevant, and find a way to book? If any of those three steps requires effort or inference, the website is losing patients who would otherwise have converted from the review stack.
AI summaries: the new layer in the research trail
An increasing proportion of patients now ask an AI assistant the question they would previously have typed into a search box. “What’s a good physio for running injuries near me?” or “Is there a physio near [suburb] that does post-surgical rehab?” The AI assistant answers by naming two or three practices rather than listing a page of links. If your practice isn’t named, you don’t exist for that patient in that moment, and there’s no page-two equivalent to drift into. Our piece on showing up in AI search as a physio practice covers the full framework, but the short version is that the same conditions that make a practice appear well in local search, consistent identity, substantive content, genuine reviews across multiple platforms, also make it legible to AI systems.
The additional consideration for AI specifically is the breadth of data sources. Many AI assistants draw local business data from Bing Places and Apple Business Connect as well as Google, which means a practice with a complete Google Business Profile but no Bing or Apple presence is invisible to a share of AI-mediated enquiries. The local search foundation that underpins Google visibility also underpins AI visibility; they’re not separate projects.
The AI search answer is winner-takes-most in a way the old results page never was. A page of ten links gave the practices ranked four through ten at least a chance. An AI summary that names two or three practices gives everyone else nothing.
What to prioritise in the next 90 days
Reviews first. They affect all three layers of the research trail simultaneously: they improve local pack visibility, add credibility on the website (via the star rating visible in search results), and contribute to the review corpus that AI systems draw on. A consistent, compliant review cultivation process is the highest-return single activity most practices could start this month.
Then specificity on the website: service pages that address specific conditions and activities rather than generic service descriptions. Then the one-time tasks: claiming and completing Bing Places and Apple Business Connect profiles, which take a few hours and extend visibility across AI search platforms that most practices haven’t considered. These are not sophisticated activities. They’re systematic ones, and the gap between practices that do them and practices that don’t is widening as more patient research moves to AI-mediated formats.
What practice owners ask about the patient research trail
How much do reviews actually matter compared to other factors?
More than most practice owners realise, and the gap is measurable. Practices with 80 or more reviews and a 4.7 or higher average consistently outperform practices with fewer or older reviews, even when other factors, location, service range, booking availability, are comparable. Reviews affect not just local pack ranking but also the confidence a patient feels before they call or book. A practice with strong reviews that a patient can read removes a significant amount of the uncertainty that otherwise leads to “I’ll think about it” behaviour.
Does my website content actually affect whether AI mentions my practice?
Yes, though the mechanism is indirect. AI systems don’t always draw from practice websites directly; they draw from the sources they’ve indexed, which include websites. Content that directly answers common patient questions, with clear headings and specific language, is more extractable by AI systems than generic service descriptions. A practice that publishes a substantive piece on managing a specific running injury under its own name is a more useful source for an AI answering a question about that injury than a practice with only a thin services page.
Should I be worried that AI might present wrong information about my practice?
It’s worth checking regularly. Ask an AI assistant the question a real patient would ask about your suburb and your main services, and see what appears. If incorrect information appears, the most reliable fix is to ensure your own website, Google Business Profile, Bing Places, and Apple Business Connect all carry clear, consistent, correct information. The more corroborated your details are across sources, the more accurately AI systems tend to represent you.
Getting found is half of marketing’s whole job.
The Build sets up the full system, your Google presence, condition pages, reviews and tracking, on tools you own. Then we hand you the keys.
See The Build →This article is general commentary for practice owners and is not legal, clinical or regulatory advice. Marketing for regulated health services must comply with the National Law and AHPRA guidance. Check the current requirements before acting.