How to get your medical practice recommended by ChatGPT

Patients now ask ChatGPT questions like "who is a good dermatologist near me for eczema" before they ever open Google. The assistant answers with two or three names. This guide explains how those names are chosen and what a practice can do to become one of them.

There is no ad slot inside ChatGPT that a practice can buy, and no form to submit. The assistant builds each answer from what it can retrieve at that moment and from what it learned in training. Your job is to control the inputs: what the retrieval step finds, and how confidently the model can repeat it. Everything below follows from that.

How does ChatGPT decide which doctors to recommend?

When a user asks a question that needs current or local information, ChatGPT runs a web search in the background, reads a handful of the returned pages, and writes an answer that cites some of them. OpenAI has said that ChatGPT search relies on third-party search providers alongside its own crawler, OAI-SearchBot, and Microsoft's Bing index is widely reported to be one of those providers. In practice, that means three layers decide whether your practice appears:

1. Retrieval. Does a search for the patient's question return your website, your Google or Bing listing, or a directory page that names you? If none of these surface, the model never sees you.
2. Extraction. Once a page is retrieved, can the model pull a clean fact out of it? "Dr. Lee is a board-certified dermatologist in Austin who treats adult eczema and psoriasis" is extractable. A hero banner that says "Your skin, our passion" is not.
3. Confidence. Health is a high-stakes topic, and assistants are tuned to be cautious. A practice that is described the same way on its site, its listings and independent sources is a safer recommendation than one with contradictory or thin information.

Training data matters too, especially for well-known institutions, but for a private practice the live retrieval layer is where most of the opportunity sits. It also changes quickly, which is good news: improvements can show up in weeks rather than the months organic search usually takes.

What sources does ChatGPT read about a practice?

Look at the citations ChatGPT shows for local healthcare questions and a pattern emerges. The sources fall into four groups:

No single source wins on its own. The practices that get named consistently tend to show up in at least three of these groups with the same core facts.

Step 1: Make sure AI crawlers can actually reach your site

This sounds obvious, yet it is the most common blocker we find. Many practice websites sit behind security plugins or CDN bot-protection rules that quietly block anything that is not a mainstream browser. Check your robots.txt file and your firewall settings for these user agents:

• OAI-SearchBot and ChatGPT-User (OpenAI search and live browsing)
• GPTBot (OpenAI training)
• PerplexityBot and Perplexity-User
• ClaudeBot, Claude-SearchBot and Claude-User (Anthropic)
• Bingbot, which feeds Microsoft Copilot and is reported to feed ChatGPT search results
• Googlebot, which feeds Google AI Overviews and AI Mode

Blocking the training crawlers is a legitimate business choice. Blocking the search and user-triggered agents means an assistant cannot open your page when a patient asks about you, which is almost never what a practice wants. Also confirm that your key content is in the HTML itself. Pages that only render text through heavy JavaScript are often read as empty by AI crawlers.

Step 2: State the facts an assistant needs, in plain sentences

Assistants quote sentences. Give them good ones. Every practice website should have, in text, on a crawlable page:

• The full practice name, every physician's name with credentials (MD, DO, board certifications) and the specialty in plain words
• The city, neighborhood and full address, plus the areas you realistically draw patients from
• The conditions you treat and the procedures you perform, each named the way patients say them
• Insurance accepted or a clear statement that you are out of network, plus telehealth availability
• Languages spoken, hours, how to book and how quickly new patients are seen

Write these as short declarative sentences near the top of the relevant page. A good test: if you copied one sentence into a text message with no context, would it still be true and useful? That is the kind of sentence an assistant can safely reuse.

Step 3: Build one page per condition or procedure

Patients ask specific questions: "best knee replacement surgeon for a younger patient", "dermatologist who treats hidradenitis". A single "Services" page with twenty bullet points gives the model nothing to cite for any of them. A dedicated page for each important condition or procedure, with a direct answer at the top, a short explanation of how your practice approaches it, who it suits, what recovery looks like and a FAQ section, gives the model a precise match for a precise question.

Research from Princeton and Georgia Tech on what the authors named Generative Engine Optimization found that adding clear statistics, quotations and citations to content increased its visibility in generated answers by as much as 40 percent in their tests. Medical content lends itself to this naturally: cite the clinical guideline you follow, give real recovery timelines and name the specialty society whose standards apply.

Step 4: Add structured data that confirms who you are

Schema markup does not guarantee anything on its own, but it removes ambiguity. At minimum, a practice site should carry:

• MedicalClinic or Physician on the home and location pages, with address, phone, geo coordinates, opening hours, medical specialty and sameAs links to your Google profile, Healthgrades, Zocdoc and social profiles
• Person or Physician for each doctor, with credentials and affiliations
• MedicalProcedure or MedicalCondition where appropriate on treatment pages
• FAQPage on pages with real question-and-answer content

The sameAs links matter more than most people realize. They tell every system reading the page that the practice on your website is the same entity as the listing on Healthgrades and the profile on Google, which is the consistency signal assistants rely on. Our schema guide for medical practices covers the exact properties.

Step 5: Fix your listings and directory profiles

Because ChatGPT search leans on external indexes and directories for local questions, your Bing Places profile deserves the same attention as your Google Business Profile. Most practices have never claimed it. Then go through the major healthcare directories and make sure every one lists the same name, address, phone number, specialty and website. Mismatched suite numbers, an old phone line or a physician who left three years ago all reduce confidence.

Reviews count as well. Assistants frequently mention review volume and rating when recommending a provider. Ask every satisfied patient for a review through a compliant, consistent process, respond to reviews professionally without disclosing any patient information, and never buy or gate reviews. The FTC's rule on fake reviews, in force since October 2024, makes that a legal issue as well as an ethical one.

Step 6: Earn mentions outside your own website

An assistant is more comfortable naming a practice that others have named. For a private practice, the realistic sources are local news interviews, hospital or university affiliation pages, specialty society directories, podcasts, guest articles in regional health publications and community Q&A. One well-placed interview in which a physician explains a treatment in plain language can be cited for years.

Step 7: Measure what the assistants actually do

Standard analytics hide most of this activity. Two measurements show whether the work is paying off:

• AI crawler visits in your server logs, separated into training crawlers, search indexers and user-triggered fetches such as ChatGPT-User. A user-triggered fetch means a real person asked an assistant something and the assistant opened your page to answer.
• Referred sessions from assistant domains such as chatgpt.com, perplexity.ai, gemini.google.com and copilot.microsoft.com. ChatGPT also appends utm_source=chatgpt.com to many outbound links, which makes these visits easy to segment in GA4.

Run a fixed set of patient-style prompts every month across ChatGPT, Perplexity, Gemini and Google AI Mode, and record whether your practice is named, how it is described and which sources are cited. The cited sources are your to-do list. See how to measure AI referrals for a full setup.

What does not work

• Buying placement. Nobody can sell a guaranteed recommendation inside ChatGPT. Anyone promising it is selling something they do not control.
• Keyword-stuffed AI content. Hundreds of thin, generic pages dilute your site and, in health, can hurt how search engines assess your expertise.
• Hidden text or prompt injection. Text aimed at manipulating AI systems is a fast way to be treated as untrustworthy, and search engines have policies against it.
• Changing nothing offline. If your phone goes to voicemail and your reviews mention long waits, assistants will read that too.

How long does it take?

In our own work, AI crawlers typically start fetching a restructured site within days of the changes going live, and the first assistant-referred visits tend to follow within two to eight weeks. Organic Google rankings for competitive local terms usually take three to six months. Results depend on your specialty, your market and how much third-party presence you already have, and no one can promise a specific timeline for a system they do not operate.

A 30-day starting checklist

This is the same sequence we follow for our own clients, and it is how our own site became a source that AI assistants cite for medical marketing questions. If you would rather have a specialist do it, our GEO service for medical practices covers each step, and the AI visibility audit is the usual starting point.

Frequently asked questions

Can I pay OpenAI to have ChatGPT recommend my practice?

No. ChatGPT answers are not for sale as placements, and any vendor promising a guaranteed recommendation is describing something it cannot control. What you can influence is whether the assistant finds your practice, understands it and trusts it enough to name it.

Does ChatGPT use Google results?

ChatGPT search uses its own crawler and third-party search providers, and Microsoft Bing is widely reported to be one of them. That is why a claimed and complete Bing Places profile and good Bing visibility matter for ChatGPT, even though most practices only manage their Google profile.

Will blocking GPTBot stop ChatGPT from recommending me?

Blocking GPTBot limits use of your content for training. Live answers depend mainly on OAI-SearchBot and ChatGPT-User. If you block those, ChatGPT cannot open your pages when a patient asks about you, which usually hurts visibility.

Do reviews affect whether AI assistants recommend a doctor?

Yes, indirectly. Assistants often cite directories and listings that display ratings and review counts, and they tend to mention them when recommending a provider. A steady flow of genuine reviews on Google, Healthgrades and similar platforms supports recommendations.

How often should I check what ChatGPT says about my practice?

Monthly is a practical rhythm. Use the same set of patient-style prompts each time, across several assistants, and log whether you are named and which sources are cited. Answers vary between runs, so trends matter more than any single result.

Keep reading