A patient with a sore shoulder used to type “shoulder pain doctor near me” into Google and pick from a map pack. Now a growing share of them ask an assistant instead: who should I see for shoulder pain that’s worse at night, and are there any good sports medicine doctors near Bellevue. The assistant answers in a paragraph and names two or three practices. If yours isn’t one of them, you don’t show up at all. There’s no page two to climb toward.
Key Takeaways
- AI assistants and AI Overviews summarize a handful of sources instead of listing ten links, so the practical goal shifts from ranking to getting cited.
- Assistants lean heavily on structured data, third-party listings, and clearly written answers, which means your Google Business Profile and directory presence matter more than they did five years ago.
- Content that answers a specific patient question in the first two sentences gets pulled into summaries far more often than content that builds slowly to a point.
- Physician credentials, hospital affiliations, and reviewed-by bylines carry real weight for health topics, where AI systems are tuned to be cautious about sourcing.
- You can check your visibility today by running your own practice’s key questions through the major assistants and writing down who gets named.
What changed, in plain terms
Traditional search gave you a list and let the patient choose. AI search does the choosing, then explains itself. The system reads a set of sources, decides which ones are credible and relevant, and writes an answer that names a few of them. Everything else is invisible, not ranked lower.
For medical practices this cuts both ways. The bad news is obvious. The good news is that these systems are unusually cautious with health questions, so they weight signals that a well-run practice can actually control: verified credentials, consistent listings, plain answers to real clinical questions, and content attributed to a named physician rather than to nobody.
The foundation is your listings, still
Assistants pull heavily from structured, verifiable sources when they answer local health questions. That means your Google Business Profile, your hospital system directory entry, your insurance network listings, Healthgrades, Vitals, WebMD Care, and your state medical board record. When those disagree with each other, the assistant has to guess, and guessing usually means picking the practice whose data is clean.
Audit the obvious fields across every listing you can find. Practice name exactly as you write it. Address with the same suite formatting. Phone number, hours including holiday hours, accepted insurance plans, and the specialties and subspecialties for each provider. Then check the individual physician listings, not just the practice ones. Assistants frequently answer at the doctor level, and a physician whose NPI record shows an old address at a practice she left in 2021 creates exactly the kind of conflict that gets a name dropped from an answer.
Write the way patients ask
Conversational search means longer, messier queries. Not “knee pain” but “my knee hurts going down stairs but not up, do I need an MRI.” Content built around the short keyword rarely matches that. Content built around the actual question does.
Answer first, explain second
Put a direct two or three sentence answer immediately under the heading, then expand. AI systems extract passages, and a passage that stands alone gets used. A passage that begins “There are several factors to consider when evaluating knee pain” doesn’t stand alone. It stalls.
Build symptom pages, not just service pages
Most practice sites have a page for each procedure. Fewer have pages for the symptoms that send people looking. A patient doesn’t search for arthroscopic meniscectomy. They search for what’s wrong with their knee. Symptom pages sit earlier in the journey and they’re where assistants find the language patients use, so they get quoted more.
Cover the logistics questions too
What insurance do you take, do you see kids, how long is the wait for a new patient appointment, do you do same-day visits, is there parking, do you offer telehealth for follow-ups. These aren’t glamorous but assistants get asked them constantly and most practice websites answer none of them in text a machine can read. A well-organized FAQ page covering practical access questions is one of the cheapest wins available.
Credibility signals for health content
Health and medical topics sit in the category search engines have long treated with extra scrutiny, and AI systems inherited that caution. A few things move the needle:
- Named physician authorship. Every clinical article should carry a byline or a “medically reviewed by” line with the physician’s full name, credentials, and a link to their bio page.
- Physician bio pages with real depth. Board certifications, residency and fellowship, hospital affiliations, languages spoken, years in practice, and the conditions they see most.
- Physician and MedicalOrganization schema connecting each provider to the practice, the specialties, and the locations served.
- Review dates on clinical content. A visible “last reviewed” date tells both patients and machines the information is maintained.
- Citations to primary sources where you’re making clinical claims, linking to specialty societies or peer-reviewed literature rather than to content mills.
None of this is exotic. It’s the same thing a referring physician would want to see, which is roughly the standard these systems are aiming at.
Reviews feed the summary
When an assistant recommends a practice, it often justifies the recommendation with a sentence drawn from patient reviews. Something like “patients frequently mention short wait times and thorough explanations.” That sentence comes from the actual text of your reviews, which gives you a reason to care about what patients write and not only how many stars they leave.
Practices that ask a specific question in their review request get more useful text back. “If you have a minute, would you mention what you came in for and how the visit went?” produces reviews that name conditions and describe the experience. Generic requests produce five stars and no words, which is fine for your average but useless as source material. Respond to reviews too, including the unhappy ones, since those replies are public text that gets read alongside everything else.
How to check where you stand
Spend forty-five minutes on this and you’ll know more than most of your competitors do. Write down fifteen questions a prospective patient might ask about your specialty and your city. Mix condition questions with practical ones. Run each through ChatGPT, Google’s AI Overviews, Perplexity, and Gemini, using a fresh session so past history doesn’t skew things.
Log which practices get named, which sources get cited, and whether anything the assistant says about you is out of date. Wrong hours and a retired physician still listed as accepting patients are both common and both fixable. Repeat the exercise quarterly. The scores will move, and the movement tells you whether your listings cleanup and content work are landing.
What not to do
Don’t publish a hundred thin AI-written condition pages in the hope of blanketing the topic. Health content is exactly where thin sourcing gets filtered out, and a page of generic symptom copy with no physician attached is worse than no page. Don’t stuff FAQ schema onto pages that have no visible FAQ. And be careful with anything that could pull protected health information into public content, including review responses, where the safe reply acknowledges the feedback and moves the conversation offline without confirming that the person was ever a patient.
The practices winning here aren’t doing anything clever. They keep their listings accurate, they let their physicians put their names on the content, and they answer the questions patients actually type. That was good practice before AI search. It just matters more now that being left out of the answer means being left out entirely.
