AI search optimization: how to get your practice recommended by ChatGPT and Google AI

AI search optimization is the work of making your practice one of the clinics that ChatGPT, Perplexity and Google AI Overviews name when a patient asks for a recommendation — and it comes down to six controllable signals: being indexed where the assistants look (including Bing), a strong review profile, presence in trusted directories, content written to be quoted, structured data, and an llms.txt file. Practices that put these in place get named; practices that only do classic SEO often stay invisible to AI answers even when they rank well on Google.

This matters now because patient behavior has already shifted: instead of scanning ten blue links, a growing share of patients asks an assistant "best dermatologist near me that takes my insurance" and gets three names. Being one of those three is the single least crowded opportunity in healthcare marketing in 2026, because most practices have not even started. This discipline is often called GEO — generative engine optimization — and here is how it actually works.

How AI assistants decide which clinics to recommend

Each assistant builds answers differently, but they draw from overlapping sources:

• ChatGPT browses through Bing's index and leans on established directories and review platforms when asked for local recommendations.
• Perplexity retrieves live web pages and cites them — it favors pages that answer the question directly and can be quoted verbatim.
• Google AI Overviews assembles answers from pages Google already trusts, with heavy weight on local signals: your Business Profile, reviews and consistent citations.

Notice what is common: none of them invent recommendations. They aggregate what the web already says about you. AI search optimization is therefore not a trick — it is making the web's record of your practice complete, consistent and quotable. The typical mistake is treating this as a separate project from SEO. It is the same foundation with a different output layer, which is why our healthcare SEO guide is the natural companion to this one.

Step 1: make sure Bing can see you

ChatGPT's browsing runs on Bing, yet most practices have never once checked their Bing presence. In the first week:

• Verify your site in Bing Webmaster Tools and submit your sitemap.
• Claim your Bing Places listing and match it exactly to your Google Business Profile.
• Search your practice name on Bing and fix whatever looks wrong or outdated.

The typical mistake: assuming Google coverage implies Bing coverage. It often does not, especially for newer sites, and an unindexed site simply cannot be recommended by ChatGPT.

Step 2: reviews are the trust signal AI leans on hardest

When an assistant justifies a recommendation, it almost always references reputation: rating, volume, and what patients say. That means your review strategy is now also your AI strategy:

• Steady velocity on Google reviews — a few every week, indefinitely.
• Reviews that mention specific services and locations, because assistants match those phrases to patient questions.
• Written responses to every review — never confirming the reviewer was a patient, which would be a HIPAA problem.

Give this 8-12 weeks of consistent work before expecting the profile to look different. Your Google Business Profile is the hub all of this feeds into, for both Google AI Overviews and the map results patients still check.

Step 3: be present in the directories assistants already trust

Ask ChatGPT for a doctor in any US city and watch where the names come from: Healthgrades, Zocdoc, Vitals, WebMD, Yelp, state medical board listings. These platforms have the authority and structure AI models trust. Claim and complete your profiles on the ones relevant to your specialty, keep name, address and phone identical everywhere, and add real photos and service descriptions. The typical mistake is a half-claimed profile with an old address — inconsistency reads as unreliability to a model deciding whether to cite you.

Step 4: publish content an AI can quote

Assistants quote pages that answer questions in complete, self-contained sentences. Rework your key pages so the first sentence under each heading directly answers the question the heading poses. Then build pages around the questions patients actually ask assistants: "does an implant hurt", "how much does a consultation cost", "what happens at the first visit". Add a genuine FAQ section to every service page, with plain-text answers of two to four sentences. The typical mistake is publishing vague brochure copy — "we offer comprehensive care" gives a model nothing to quote, while "a first orthodontic consultation takes about 45 minutes and includes a panoramic X-ray" gives it everything. Two more habits that raise quotability: name the doctor and their credentials on every service page, because assistants prefer attributable medical information, and keep pages current — a model choosing between two clinics will lean toward the one whose content reflects this year, not 2021. Plan the first 4-6 weeks for reworking your top ten pages, then one new question-page per week.

Step 5: structured data and llms.txt

Two technical pieces close the loop:

• Schema markup: Physician or MedicalClinic schema on your homepage and location pages, FAQPage schema on your FAQs, plus opening hours, geo-coordinates and services. This is how machines confirm who you are, what you treat and where.
• llms.txt: a plain-text file at yourdomain.com/llms.txt that summarizes your practice, specialties, locations and key pages for AI crawlers. It is an emerging standard, it costs an hour to create, and it puts you ahead of virtually every competitor.

The typical mistake is installing a schema plugin and never checking the output — validate it, because broken markup is ignored.

Step 6: measure it, because this is now measurable

Ask ChatGPT, Perplexity and Google the recommendation questions your patients ask — monthly, in a clean session — and log whether you appear and what the assistant says about you. A simple tracking sheet works:

• Ten prompts patients would realistically use ("best [specialty] in [city]", "[procedure] near [neighborhood] that takes [insurance]").
• For each: do you appear, in what position, and which source does the assistant cite.
• Repeat monthly and watch the trend, not any single answer.

Check your analytics for referrals from chatgpt.com and perplexity.ai, and ask new patients how they found you; "ChatGPT recommended you" is an answer front desks now hear weekly. Expect first appearances within 2-3 months of doing the steps above, faster in less competitive specialties. The typical mistake is testing once from your own logged-in account, getting a flattering answer skewed by your history, and declaring victory.

How Medical Marketing helps

Medical Marketing implements this entire GEO stack for client practices — Bing indexing, review engines, directory cleanup, citable content, schema and llms.txt — as part of the patient acquisition systems we have refined over more than 10 years and 10+ million euros invested for thousands of clinics and doctors. As a medical marketing agency serving the US market, we can audit how AI assistants currently talk about your practice and show you the gaps in a free 30-minute consultation.

Frequently asked questions

What is AI search optimization for a medical practice?

It is the process of making your practice visible and recommendable to AI assistants like ChatGPT, Perplexity and Google AI Overviews. It combines classic local SEO signals — reviews, directories, consistent listings — with machine-readable elements like structured data, llms.txt and content written in complete, quotable answers.

How does ChatGPT choose which doctors to recommend?

ChatGPT browses through Bing's index and aggregates what trusted sources say: review platforms, healthcare directories like Healthgrades and Zocdoc, and your own website. Practices with strong, consistent profiles across those sources get named; practices that are absent from them rarely appear in answers.

How long does it take to show up in AI recommendations?

In our experience, practices that fix Bing indexing, directories, reviews and structured data start appearing in some AI answers within 2-3 months. Competitive specialties in large metros take longer because assistants have more well-documented alternatives to choose from.

Is GEO different from SEO?

They share the same foundation — authority, reviews, consistent local signals — but GEO adds an output layer: content structured as self-contained answers, schema markup, llms.txt and presence in the specific sources each assistant retrieves from, including Bing. Good SEO is necessary but no longer sufficient on its own.

Does AI search optimization work for small practices?

Yes, and right now it favors them. Because most clinics have not started, a small practice that completes its directories, builds steady reviews and publishes quotable content can outrank much larger competitors in AI answers. The window narrows as adoption grows, which is the argument for starting early.