ChatGPT SEO is the practice of making your website retrievable, verifiable, and quotable enough that ChatGPT cites your practice when a patient asks it a question. It works differently from Google SEO because ChatGPT does not rank pages, it selects sources: it runs a live search, pulls a small set of pages, and cites the ones that answer the question cleanly. If your practice is not in that small set, you are invisible to a channel that now reaches more than 800 million weekly users.
I come at this from a specific angle. I spent 20 years as a software engineer, most of it building large-scale systems, including a platform that grew from zero to $500M. I have also been a functional medicine patient for 15 years. So when I watched patients start asking ChatGPT "who should I see for my thyroid issues" instead of typing it into Google, I did what an engineer does: I traced the retrieval pipeline to figure out where a practice can actually intervene. This article is that trace, written for practice owners.
The search data backs up the urgency. "ChatGPT SEO" gets about 1,300 searches a month with a competition score of 0.12, which is close to nothing. "How to rank in ChatGPT" sits around 260 searches a month and has risen all year. Your competitors are not working on this yet. That window will not stay open.
One scope note before we start. This article is tactical and ChatGPT-specific. If you want the full operating playbook across every AI surface (Google AI Overviews, Perplexity, Gemini, Copilot), including the audit and measurement system, that lives in AI Search Optimization for Medical Practices: The Playbook. Read this one first if ChatGPT is where you want to win.
How ChatGPT actually finds and cites sources
ChatGPT answers about your practice come from two separate systems: the model's training data, which is a frozen snapshot of the web, and live search retrieval, which fires when a question needs current or specific information. Local health questions almost always trigger live retrieval, because "best functional medicine doctor near Scottsdale" is exactly the kind of query the model knows it cannot answer from memory. That means your ChatGPT visibility is mostly a retrieval problem, not a training-data problem, and retrieval problems are fixable in weeks, not years.
Here is the pipeline as an engineer sees it:
- Query rewrite. The user asks something conversational ("who's good for hormone therapy in Denver that isn't a chain"). ChatGPT rewrites it into one or more search-engine-style queries.
- Live web search. Those rewritten queries hit a search index. ChatGPT's search layer has historically leaned on Bing's index, so if Bing cannot see your site, ChatGPT often cannot either. This is the single most overlooked fact in this entire topic.
- Page fetch and extraction. ChatGPT pulls a handful of results, typically well under ten, and reads them. Pages that answer the question in clear, extractable prose get used. Pages that bury the answer under a hero video and three paragraphs of "welcome to our healing journey" get skipped.
- Synthesis with citations. The model composes an answer and cites the pages it leaned on. Being fetched is not enough. You have to be the page worth quoting.
Two practical consequences fall out of this. First, the citation set is tiny. Ranking eleventh in Bing for a query is functionally the same as not existing, in a way that is even more brutal than Google's page two. Second, extraction quality matters as much as ranking. A page that ranks fifth but states the answer in the first two sentences can get cited over a page that ranks second but rambles.
There is a third path worth knowing: ChatGPT also draws on what other sites say about you. Directory listings, "best of" articles, Reddit threads, and local news coverage all feed the same index. For recommendation-style questions ("best med spa in Austin"), ChatGPT frequently cites roundup articles rather than individual practice sites. You want to be in those roundups. More on that below.
Test your practice's visibility before you change anything
Before touching your website, spend 30 minutes running real patient prompts through ChatGPT and recording what comes back. This baseline tells you which of three problems you have: ChatGPT has never heard of you, ChatGPT knows you exist but does not recommend you, or ChatGPT describes you inaccurately. Each problem has a different fix, and most practices guess wrong about which one they have.
Open ChatGPT (make sure web search is active, which it is by default for these query types) and run prompts in these categories, substituting your specialty and city:
| Prompt type | Example prompt | What it tests |
|---|---|---|
| Direct brand | "What can you tell me about [Practice Name] in [City]?" | Whether your entity exists in the index at all |
| Local recommendation | "Best functional medicine doctor in [City] for thyroid problems" | Whether you make the citation set for money queries |
| Condition + modality | "Where can I get NAD IV therapy in [City] and what does it cost?" | Whether your service pages are extractable |
| Comparison | "Is [Practice Name] or [Competitor] better for bioidentical hormone therapy?" | What sources ChatGPT trusts about you versus them |
| Skeptic | "Is [Practice Name] legit? Any red flags?" | Which reviews and third-party mentions surface |
| Logistics | "Does [Practice Name] take insurance? How much is the initial visit?" | Whether your practical info is published anywhere readable |
Record three things per prompt: was your practice mentioned, what sources were cited, and was the information accurate. Then run the same prompts for your two strongest local competitors and note whose pages get cited. When a patient says "ChatGPT sent me," and they will, you want to know which prompt and which page did the work. The full measurement protocol, the same one I recommend for every AI surface, lives in the AI search optimization playbook.
Getting into ChatGPT's citation set
Getting cited comes down to three layers, in order: be indexable by the search layer ChatGPT uses, be the clearest answer on the page for the questions patients actually ask, and be corroborated by third-party sources the model trusts. Most practices fail at layer one without knowing it, which makes everything else irrelevant.
Layer 1: Make sure Bing can see you
Google Search Console is standard for practices. Bing Webmaster Tools almost never is, and that is the gap. Set it up this week:
- Create a Bing Webmaster Tools account and verify your site. You can import your settings directly from Google Search Console in about five minutes.
- Submit your sitemap and check the index coverage report. It is common to find practice sites with a third of their service pages absent from Bing's index while fully indexed in Google.
- Check your robots.txt for blocks on
bingbot. Also decide deliberately whether you allowOAI-SearchBot(OpenAI's search crawler) andChatGPT-User(the agent that fetches pages during a live conversation). If your goal is to be cited, these should not be blocked. Some security plugins and CDN bot filters block them by default. - Fix crawl errors Bing reports even if Google reports none. The two indexes disagree more than people assume.
Layer 2: Structure pages so the answer is extractable
ChatGPT quotes pages that state answers plainly. For every service you offer, the page should answer the patient's actual question in the first two or three sentences under the heading, before any elaboration. Not "At Vitality Wellness, we believe in a root-cause approach." Instead: "A functional medicine initial consultation at our Denver clinic is 75 minutes, costs $395, and includes a comprehensive lab review. We do not bill insurance directly, but we provide superbills."
Concretely:
- One page per service, one page per condition you treat. A single "Services" page listing twelve modalities gives the model nothing specific to retrieve for any of them.
- Publish prices, or at least ranges. "Cost" questions are among the most common patient prompts, and ChatGPT will cite whoever publishes numbers. If your competitors hide pricing and you publish "$250 to $400 depending on labs," you win those citations by default.
- Use headings that mirror real questions. "How much does semaglutide cost at our clinic" beats "Investment" as a heading, every time.
- Add Physician, MedicalClinic, and LocalBusiness schema with consistent name, address, phone, and provider credentials. Schema does not directly cause citations, but it removes ambiguity about who you are, which matters for the skeptic and brand prompts.
- Put provider credentials in crawlable text. Board certifications, training (IFM, A4M, and similar), years in practice. Health queries get extra scrutiny from these systems, and pages with verifiable author credentials are safer for the model to cite. This is the same E-E-A-T logic I covered in the complete healthcare SEO guide, applied to a new retriever.
The structural discipline here overlaps heavily with answer engine optimization as a whole. If you want the full content-structure system, the AEO for healthcare guide goes deeper than I can here.
Layer 3: Get corroborated by sources ChatGPT already trusts
For "best X in [city]" prompts, ChatGPT usually cites roundup and directory content, not individual practice homepages. That means part of ChatGPT SEO happens off your website:
- Get listed in every roundup that already ranks. Search Bing for "best [your specialty] [your city]" and note which articles appear. Contact the publishers. Many local roundups accept submissions or respond to a short pitch with your differentiators and outcomes.
- Clean up your directory presence. Google Business Profile, Healthgrades, Zocdoc if relevant, Yelp, and your state association directory. Identical name, address, phone, and specialty descriptions everywhere. Inconsistent data makes the model hedge or skip you.
- Reviews are retrieval fodder. Skeptic prompts surface review content directly. A steady stream of detailed Google reviews that mention specific conditions ("helped me get my Hashimoto's under control") gives ChatGPT quotable third-party evidence. Ask for reviews systematically, and ask patients to mention what they came in for, within what review platforms allow.
- Local press and podcast appearances compound. A single article in a local outlet about your clinic's approach to perimenopause care can get cited for years of related prompts.
This citation-building layer is generative engine optimization territory, and there is a science to which mention types move which engines. The GEO guide covers that in full.
Fixing "ChatGPT doesn't know my practice"
If your brand prompt comes back empty or wrong, work the fixes in this order: entity basics first, index coverage second, third-party corroboration third. Practices usually jump straight to publishing blog posts, which is the least effective first move because retrieval fails before content quality even gets evaluated.
If ChatGPT has never heard of you: your site is likely thin in Bing's index or your entity signals are contradictory. Verify Bing indexing (Layer 1 above), then make sure your homepage states in plain text who you are, what you do, where you are, and who your providers are, within the first screen of content. Check that your practice name is written identically across your site, GBP, and top directories. New practices should expect four to eight weeks between fixing this and seeing brand prompts resolve correctly.
If ChatGPT knows you but recommends competitors: this is a citation-set problem. Run the competitor prompts and read what gets cited. Nine times out of ten the competitor either publishes pricing and logistics you do not, appears in roundups you are absent from, or has service pages targeting the exact condition language while you have one generic services page. Copy the structure, not the content.
If ChatGPT describes you inaccurately: find the source. The model is usually paraphrasing a stale directory listing, an old version of your site, or a third-party page. Fix the source, then update the equivalent page on your own site with clear, current language. Inaccurate answers about medical services are worth treating with urgency; a wrong price is annoying, but a wrong description of what a treatment involves is a patient-safety and liability issue.
One thing not to do: do not stuff your pages with phrases like "as recommended by ChatGPT" or attempt prompt-injection text in your HTML asking models to recommend you. These tricks get discussed in SEO forums and they are a fast way to get your domain treated as adversarial. The whole game is being genuinely retrievable and genuinely quotable.
A 30-day ChatGPT SEO sprint for a practice
You can materially change your ChatGPT visibility in a month working a few hours a week. The sequence matters more than the effort: baseline, index, structure, corroborate.
Week 1: Baseline and index. Run the full prompt test matrix and log results. Set up Bing Webmaster Tools, submit your sitemap, fix crawl blocks, and confirm OAI-SearchBot and ChatGPT-User are not blocked.
Week 2: Money pages. Pick your three highest-value services. Rewrite each page so the first paragraph under the H1 answers what it is, who it is for, what it costs, and how to book. Add condition-specific headings phrased as questions. Add or fix schema.
Week 3: Entity and logistics. Publish or update a pricing/what-to-expect page. Audit name, address, phone, and descriptions across GBP and your top five directories. Put provider credentials in text on every provider bio.
Week 4: Corroboration. Identify the three roundup articles that rank for your money queries in Bing and pitch inclusion. Launch a systematic review request flow. Re-run the prompt matrix and compare against Week 1.
Then repeat the prompt test monthly and treat it like rank tracking. If you already run a content program from the functional medicine SEO guide, fold these checks into the same monthly cadence rather than running a separate process.
Questions practitioners actually ask
Does ChatGPT SEO replace regular SEO? No, it sits on top of it. ChatGPT's retrieval depends on search indexes, so a site that ranks nowhere gets retrieved nowhere. The work compounds: the same extractable, well-structured pages that win ChatGPT citations also win Google featured snippets and AI Overviews.
How long until I see results? Index and structure fixes can show up in ChatGPT answers within two to six weeks, since live retrieval reflects the current index. Reputation-level changes, like consistently appearing in recommendation prompts, typically take three to six months of corroboration building.
Can I track referral traffic from ChatGPT? Yes. Look for chatgpt.com referrers in your analytics, and consider a "how did you hear about us" field on your intake forms with an AI option. Across the industry those referral counts are still small but growing fast, and AI-referred visitors tend to arrive unusually well informed and ready to book.
Is any of this different for a med spa versus a functional medicine clinic? The mechanics are identical. The difference is query mix: med spa prompts skew toward price and comparison ("Botox cost in [city]"), while functional medicine prompts skew toward condition and philosophy ("doctor who will actually run a full thyroid panel"). Structure your pages around your specialty's actual prompt patterns from your Week 1 baseline.
ChatGPT SEO is unusual in one respect: the technical bar is low and almost nobody in healthcare has cleared it yet. A practice that sets up Bing indexing, restructures three pages, and lands two roundup mentions is ahead of 95 percent of its market. If you would rather have this run for you as part of a full patient-acquisition system, that is what we do at Health Biz Scale, but everything you need to do it yourself is on this page. Start with the prompt test. What you find will make the rest of the work obvious.