About this update
A prospective patient asked an AI assistant for a clinic recommendation this week. Either yours was one of the two or three names in the answer, or a competitor’s was. This guide explains how those names get chosen, how to see exactly where your clinic stands today, and the work, none of it magic, that changes the answer over the following months.
How AI assistants choose which clinics to recommend
When someone asks ChatGPT, Gemini or Perplexity to recommend a private clinic, the assistant composes an answer from two sources: what it learned about the world during training, and what it can look up live, which increasingly means running a search and reading the results before replying. Google’s AI Overviews work from the live side almost entirely.
Both routes reward the same clinic. Training favours providers mentioned accurately and repeatedly across independent sources over time. Live retrieval favours what strong search visibility always favoured: pages that answer the question directly, structured data confirming who you are, a healthy Business Profile, and reviews corroborating the claims. The assistants differ at the margins (Perplexity leans hardest on citable sources, ChatGPT weighs its browsing results, Google’s Overviews draw heavily on pages already ranking), but a clinic built to be verifiable travels well across all of them.
One mechanical detail worth acting on: most AI crawlers read raw pages and do not execute the scripts some website builders rely on. If your treatment content only appears after code runs, the assistants may be reading a fraction of the site your patients see. Ask your developer to confirm the substance is visible without scripts.
First, measure: the five-prompt test
Before changing anything, establish the baseline no ranking guide tells you to take. Run each of these in ChatGPT, one other assistant, and Google (noting any AI Overview), substituting your details.
- Best [your specialty] clinic in [your town or city].
- Who is the best private [your specialty] consultant near [your area]?
- Is [your clinic name] good? What do reviews say?
- How much does [your main treatment] cost in [your city], and where should I go?
- I have [a symptom your patients present with]. Which private clinics in [your city] should I consider?
Record who is named, whether you are, what is said and whether it is accurate. A spreadsheet with one row per prompt and one column per month is the whole system. Repeat monthly, because AI answers drift: sources get re-weighted, competitors publish, and a clinic cited in March can vanish by June without anyone noticing unless someone is checking. That drift is also the honest reason this work is a routine, not a project.
Decide which questions you want to own
The strategic shift AI search demands is from keywords to questions. A clinic cannot be the answer to everything; it can absolutely become the answer the assistants reach for on three to five specific patient questions in its catchment: the symptom question, the treatment-choice question, the cost question, the who-is-best question. Write those prompts down. Every piece of work that follows should be traceable to one of them, and your monthly test measures exactly those. Owning a handful of prompts deeply beats being vaguely present across dozens, because assistants give two or three names, not ten blue links.
The playbook: what actually moves the answer
1. One verifiable identity
Your clinic’s name, address, phone number and description must match across your website, Google Business Profile, directories such as Doctify, professional registers and social profiles. Add organisation, medical business and FAQ structured data to the site so the machines have an identity card to check against. Inconsistency reads as uncertainty, and uncertain providers do not get recommended.
2. Pages that answer whole questions
Rebuild your key treatment and pricing pages around the questions patients actually ask, with question-shaped headings and direct, clinically reviewed answers beneath them. Prefer a clear comparison table to a decorative image; assistants quote tables and skip graphics. This is the same work that wins featured snippets and AI Overviews, which is why it repays itself twice.
3. Reviews the machines can quote
Assistants read review volume, recency and sentiment, and paraphrase the gist when describing your clinic. Build a steady review routine, and respond professionally to everything, remembering that confidentiality still applies: never confirm someone was a patient or discuss their care in a reply. Your responses are now part of what the machines summarise about you.
4. Independent mentions
Training data is built from the wider web, so citations beyond your own site carry weight even without links: directory profiles, expert commentary in the press, professional body listings, and genuine community discussion. This is the slowest lever and the most defensible one, because a competitor cannot copy your third-party footprint the way they can copy a page.
5. Measurement, monthly
Close the loop: the five-prompt test each month, alongside enquiries and bookings, tells you whether the answer is moving and catches drift early. We run a forty-prompt version of this for clients; five is enough to self-serve.
A 90-day sequence for a busy clinic
Month one: run the baseline test, fix naming consistency everywhere, ship organisation and FAQ schema, and confirm the site is readable without scripts. Month two: rebuild the two or three treatment pages behind your chosen prompts into question-led, clinically reviewed form, and start the review routine. Month three: begin the citation work (directories, one piece of expert commentary, professional listings), re-run the prompts, and compare against the baseline. Modest movement inside a quarter is realistic in most specialties; the compounding comes after, on the same slow-then-sudden curve as the SEO it is built on, and our healthcare SEO timeline guide sets those expectations honestly.
What nobody can do (and what to say to anyone who claims otherwise)
There is no submission form, no paid inclusion and no privileged access to ChatGPT’s recommendations or Google’s AI Overviews. Telling ChatGPT about your clinic in a conversation changes that conversation, not the model. Answers vary by prompt, day and user, so guaranteed placement is a promise nobody can keep, and in our view it is the clearest red flag in this young market. A credible provider shows you a testing method, the prompts tracked, the share of answers over time, and the fixes and citations shipped to move it. If you are evaluating help with this, our guide to choosing a healthcare SEO agency includes the questions that expose the difference.
When the AI gets your clinic wrong
Assume errors will happen: superseded prices quoted, a treatment attributed that you do not offer, your clinic confused with a similarly named provider. You cannot edit the assistant, but you control everything it reads. When the monthly test surfaces an inaccuracy, trace it to the likely source (an outdated directory entry, an old page, a stale third-party article), correct it there, and note the date; answers typically follow the sources within weeks. For regulated providers there is a sharper point underneath: everything you publish remains yours under the CAP Code and professional guidance, and accurate, well-evidenced source material is also your best protection against being misquoted by machines summarising it.
Work with Pulse Digital Health
Pulse Digital Health is a healthcare-only digital marketing agency working with private clinics, consultants and healthcare brands across the UK. AI visibility is built into how we work: monthly prompt audits, the identity, content and citation foundations this playbook describes, and reporting in enquiries, booked appointments and cost per new patient.
If you are looking for a trusted partner in healthcare, we would be glad to talk. Book a free discovery call and we will run the five-prompt test on your clinic live, show you who the assistants currently recommend in your specialty, and tell you honestly what it would take to change the answer.
References
- Google Search Central, AI features and your website
- Google Search Central, Creating helpful, reliable, people-first content
- Schema.org, Getting started with structured data
- Google Business Profile Help, Improve your local ranking on Google
- CAP Code, Section 12: Medicines, medical devices, health-related products and beauty products
Frequently asked questions
How do I get my business recommended by ChatGPT?
Make it easy to find, verify and corroborate: one consistent identity everywhere, structured data on your site, pages that answer real questions directly, genuine reviews, and independent mentions in directories and the press. There is no submission route; presence is earned through verifiable signals.
Can I pay to appear in ChatGPT's recommendations?
No. There is no paid inclusion in ChatGPT’s organic recommendations or Google’s AI Overviews, and no privileged submission process. Treat guaranteed-placement offers as a warning sign; credible providers offer a testing method and earned improvements, not certainty.
Does telling ChatGPT about my clinic make it recommend us?
No. What you type in a conversation shapes that conversation only; it does not update the model or influence other users’ answers. Recommendations come from training data and live retrieval of public sources, which is where the work belongs.
How do I see what ChatGPT says about my clinic?
Ask it the questions your patients ask: best clinic for your specialty in your city, whether your clinic is good, where a treatment costs what. Record who is named and what is said, check one other assistant and Google’s AI Overview, and repeat monthly to catch drift.
How long does it take to appear in AI recommendations?
Identity and structured-data fixes can be reflected within weeks; building the reviews and independent citations that drive recommendations typically takes months, on a similar curve to SEO. Expect modest movement inside a quarter and compounding afterwards, tracked through monthly prompts.
Do Google reviews affect ChatGPT recommendations?
Yes. Assistants read review volume, recency and sentiment across public sources and paraphrase them when describing providers, and reviews feed the local signals AI Overviews lean on. A steady, professionally answered review profile is one of the strongest inputs a clinic controls.
Does schema markup help AI visibility?
Yes. Organisation, medical business and FAQ structured data give assistants a machine-readable statement of who you are, what you do and where, which supports confident citation. It is one of the cheapest fixes on the playbook and most clinic sites lack it.
Why does ChatGPT recommend my competitor and not my clinic?
Usually because the competitor is easier to verify: more consistent public information, stronger reviews, clearer answer-shaped pages or more independent mentions. Run the five-prompt test to see what is said about both, then close the specific gaps rather than guessing.
Are AI recommendations accurate?
Not always. Assistants quote outdated prices, confuse similarly named providers and occasionally attribute services incorrectly. You cannot edit the answer, but correcting the sources it reads (your site, profiles, directories) typically corrects the answer within weeks, which is why monthly monitoring matters.
Is optimising for ChatGPT different from SEO?
It is an extension rather than a replacement: the same foundations of accurate content, verifiable identity and genuine authority drive both, with added emphasis on unlinked mentions, extractable formats and prompt-level measurement. Our SEO, GEO and AEO guide explains the terms in plain English.
