What Google’s new AI search guide means for your medical practice

The last time you searched online for a restaurant, a contractor or a specialist outside your network, something new appeared at the top of your results: a generated answer summarizing your options before you ever reached the first link.

If you’re like most practice owners, you probably wonder how Google generates those answers and whether your practice shows up in them.

Your patients are wondering the same thing about you, because more and more of them are using artificial intelligence (AI) to find a doctor. What that means for your practice depends largely on what your website says, and what it doesn’t.

How are patients using AI to find and choose physicians?

Patients have never had more tools to find and evaluate a physician before they make a call. For most of the past two decades, that search began with a list of blue links and review sites. Today, it increasingly begins with a generated answer.

According to a 2025 survey published by Medical Economics, more than a quarter of patients say an AI recommendation directly influenced which physician they chose, nearly equal to the influence of primary care referrals. For independent practices competing without a hospital brand behind them, that shift has real consequences for how new patients arrive and who they choose.

In May, Google published “Optimizing Your Website for Generative AI Features on Google Search,” the first time the company put all of its AI search guidance in one place. What it says, and what it doesn’t say, is worth a close read.

Is following Google’s AI search advice enough for medical practices?

Partly. The guide confirms that traditional search fundamentals still apply: solid technical structure, fast load times and accurate local business information. It also reinforces EEAT — expertise, experience, authoritativeness and trustworthiness — as the framework for building credible content.

In most industries, EEAT separates credible sources from unreliable ones. In medicine, it’s table stakes. Every physician went to medical school. Every physician has years of clinical experience. Every physician holds board certification.

When every credible option clears the same bar, EEAT stops being a differentiator and becomes a floor, and floors are still necessary. What gets a physician into AI results isn’t being credible. It’s being the right fit for a specific patient’s situation. That’s a different question, and most practice websites aren’t written to answer it.

What should a physician’s website say to show up in AI search results?

A bio written for traditional search might read: “Dr. Sarah Chen has 20 years of experience in family medicine. She is certified by the American Board of Family Medicine and completed her residency at [hospital name].”

The same bio written for AI search might read: “Dr. Sarah Chen has spent 20 years in family medicine focused on patients managing several chronic conditions at once. Her approach treats diabetes, hypertension and anxiety as interconnected problems rather than separate diagnoses, which tends to serve patients who want more than a quick visit and a referral.”

The goal isn’t to replace credential-listing but to build on it. Credentials establish legitimacy. Context creates relevance.

Both versions are accurate. Only the second gives an AI model something to work with when a patient asks, “Who is the best primary care doctor for someone managing diabetes and anxiety in [city]?” AI models surface physicians whose websites explain not just what they do but who they’re best suited to help. When that context is missing, every practice looks the same.

What does AI search actually look for when recommending a physician?

Working with a functional medicine clinic in Indiana, we ran into a situation that forced a different approach. The practice had significant website limitations: slow load times, bloated code and pages that were difficult to optimize through conventional means. By traditional search measures, it was a weak performer, with only 32 patient reviews compared with competitors that had several hundred.

Rather than fight the technical constraints, we focused on helping AI models understand who this physician was and why his background mattered to specific patients. We updated his biography, home page and condition-specific pages to explain his clinical philosophy rather than list credentials, added FAQ sections using direct-answer formatting, and rewrote blog content to reflect the way patients ask questions rather than the way practices describe services.

To understand what the AI models were actually weighing, we asked them directly, prompting ChatGPT and Claude to explain their reasoning after each search. One early response described a recommended practice as “appearing reputable and relevant.” When we pushed back and asked what specifically on the website supported that, the answer changed.

The AI pointed to a paragraph explaining the physician’s approach to patients managing overlapping chronic conditions. Not the credential list, not the review count. A paragraph about clinical philosophy was doing more work than everything else on the page combined.

Same story with competitors. Practices the AI recommended with confidence had content explaining who they were for and how they practiced. Practices it described vaguely had credential lists with nothing connecting them to a specific patient need.

Can updating website content really improve a practice’s AI search visibility?

Before the changes, the practice showed up in about 27% of 22 AI search phrases we tested across ChatGPT and Claude. Two weeks after publishing the updated content, it appeared in the top three recommendations for 82% of the same phrases. No new reviews, no paid advertising, no technical fixes. The only change was content that gave AI models a specific reason to recommend this physician over competitors with far more reviews and technically stronger websites.

We’ve seen the same pattern across family medicine, internal medicine and specialty practices. AI models recommend the physician whose website gives them a reason to, not simply the one with the longest credential list.

How can independent practices improve their AI search visibility today?

Start by reading your own bio and home page with one question in mind: If a patient described their health situation to an AI and asked for a recommendation, does anything on this page give the AI a reason to choose your practice specifically?

If the answer is no, begin with your clinical philosophy. What kinds of patients are you best equipped to treat and why? If you practice evidence-informed medicine, say so and link to the evidence. If you take longer appointments with complex patients, say that.

Format matters too. AI tools extract information most reliably from direct-answer structures, and an FAQ section where questions mirror the way patients actually search is one of the highest-leverage changes a practice can make right now. (Related: What physicians need to know about marketing)

Google’s guide calls this being “people first.” Put simply, it means treating your web presence like an introduction to a referred patient rather than a curriculum vitae.

What is the future of AI search for patient acquisition?

Google’s guide describes the emergence of AI agents capable of taking actions on behalf of users, including completing bookings, directly within search results. Nobody knows exactly when that will become routine in health care scheduling. But the trajectory is obvious to anyone watching.

Practices that give AI models a genuine reason to recommend them aren’t just winning in search results today. They’re building something that will hold up however this technology continues to develop. The physicians who start that work now won’t be scrambling to catch up later.

Michael Funkhouser is the founder of Practice Growth Co., a patient acquisition agency that works exclusively with specialty clinics and independent practices on paid advertising, search strategy and conversion optimization. He has partnered with practices across five countries.