Patients Are Asking AI Where to Go

A patient who asks an AI assistant for a recommendation does not get a page of results. They get three or four names. The assistant chose those names using signals most practices have never heard of — and the practices that get named are rarely the biggest. They are the ones the machine could read, understand, and trust.

That is the whole opportunity, and the whole risk. Healthcare is a high-trust category, which means AI assistants are unusually careful about who they recommend. They strongly prefer entities they can verify: named clinicians with credentials, a real location, specific services, and plain-language answers to the questions patients actually ask. Most independent practices have every one of those things. Very few serve them in a form a machine can use.

Why practices are unusually well-positioned — and unusually invisible

Compare a practice to a national chain. The chain has scale; the practice has something an assistant values more when the question is “who should I see”: identifiable people. A named doctor with a stated specialty, a bio, a history, and content attached to their name is exactly the kind of entity a model can describe with confidence. A brand name attached to two hundred interchangeable locations is not.

The problem is delivery. In the audits we run, the typical practice website fails the machine on three fronts at once:

• The site is built in JavaScript the crawler never runs. AI retrieval fetches raw HTML. If the page arrives empty until a browser assembles it, the assistant reads nothing — no doctors, no services, no location.
• The people are invisible in the markup. “Meet the Team” is a page of photos. There is no structured data saying this is a Physician, this is her specialty, this is where she practices.
• The service pages describe vibes instead of answering questions. “Compassionate, personalized care” cannot be quoted in answer to “how much does LASIK cost and do you offer financing?”

The five signals that decide whether a practice gets named

1. Crawlable pages. Every public page serves its full content as real HTML, with one clear title, description, and canonical URL. This is the gate. Nothing below matters until an assistant can read the site without executing code. Our playbook shows how to test this in thirty seconds.

2. Provider entities. Each clinician marked up with name, credentials, specialty, and the practice they belong to — so the assistant can resolve “Dr. Jane Doe, optometrist, Boise” as one unambiguous entity rather than a name it has seen in a dozen unrelated places.

3. Answerable content. Real answers to real patient questions: what it costs, what insurance is accepted, whether financing is available, what the first visit involves, how long recovery takes. Written in plain language on pages a machine can lift accurately, and marked up as FAQ where it genuinely is one.

4. Local corroboration. A complete Google Business Profile, consistent name/address/ phone across directories, and reviews that mention the services and the people. Assistants cross-check what your site claims against what the rest of the web says about you.

5. Named expertise in public. Articles, interviews, talks, and papers attached to the providers by name. This is the signal practices underuse most, and the one that compounds: every piece of provider-attributed content teaches the model who your clinicians are and what they know.

The compounding move: publish what patients ask

Every question the front desk answers ten times a week is an article an AI assistant can cite. “Do you finance implants?” “What is the difference between an OTC hearing aid and a prescription one?” “How do I know if I am a LASIK candidate?” A practice that answers those questions in public, under a named clinician, is building the exact material assistants look for when they decide who to recommend.

Two live examples from the Practice Mindshare series: Dr. Ryan Beck on what growing to nine locations actually took, and Dr. Douglas Beck on the future of hearing care. Both are provider-attributed, both answer questions patients and peers actually ask, and both are readable by every AI crawler that visits.

What not to do

Do not stuff pages with condition keywords, do not fabricate reviews, and do not buy a service that promises guaranteed placement in AI answers — no one controls that, and anyone claiming otherwise is selling something they cannot deliver. For a practice, the stakes are higher than for most businesses: a patient who is misled by inflated claims is a compliance problem, not just a marketing one. The honest version of this work is slower and it holds up.

Where to start

Find out what an AI crawler sees on your site today. If you would rather read the method first, the definitional explainer covers how assistants actually decide. The productized version — audit, implementation roadmap, and ongoing monitoring — is the AI Visibility Package on our Growth Packages page.