When a practice worries about AI search, the conversation is almost always about the website. That is the asset they own, so it is the asset they think about. Google's own guidance suggests it is not where the local answer mostly comes from.

What the guidance says for local

In the section on local business and ecommerce details, Google says generative AI responses can include product listings, product information and information about local businesses, and that using products like Merchant Center and Google Business Profile can help products and services be visible in both AI responses and other Search results.

It also mentions Business Agent, which it describes as a conversational experience on Google Search that lets customers chat with a brand.

None of that says the website does not matter. The same guide is emphatic that unique, non-commodity content is the largest long-run factor. But for the specific question a local patient asks, the structured local record is being named first.

The part outside your control

Google is also explicit that these features can show what is being said about products and services across the web, including in blogs, videos and forum discussions. For a practice that means reviews, local forums and third-party directory listings are candidate material for the answer a patient reads.

This is where the honest position has to be uncertainty. Nobody outside Google can see which sources were retrieved for a given answer, and Search Console reports that a page appeared without saying what was asked. So the accurate statement is that these surfaces are in scope, not that they are weighted a particular way.

The thing almost nobody does

Ask the question a patient would ask, read the answer, and write down what it said about you and what it cited. Do it for the five queries that actually matter in your catchment.

It is a twenty minute exercise and it is unreasonably informative. Practices discover a departed clinician still named, an old address, a price that has not been current for two years, a treatment listed that they stopped offering, or a competitor described more specifically than they are.

It is a sample rather than a measurement, and it should be written down as one: the query, the date, the surface, what it said. Repeated monthly it becomes something you can reason about. Done once and quoted forever it becomes another confident claim in a category already full of them.

Where the effort goes

The Business Profile is the highest-leverage thing most practices are neglecting, because it is treated as a set-up task rather than a maintained record. Services, hours, the practitioner list and the photographs decay quietly, and unlike a website nobody is looking at it.

Reviews matter here in a way that is different from their usual reputational role. They are text about your practice that the system can read. A body of specific reviews mentioning actual treatments and actual outcomes gives an answer engine something concrete to work with, which generic five-star ratings do not.

Then the website, which is where the answers to the considered questions live: what the treatment involves, what it costs, who performs it, what recovery looks like. That is the non-commodity content the guidance keeps pointing at, and for a practice it is mostly just writing down what the consultation already covers.

What this means for an operator

Run the five queries a patient in your catchment would actually type, read what the AI answer says about your practice, and log it with the date. Fix whatever is factually wrong in the Business Profile first, because that is the record you control and nobody is checking.