Guide · Tracking
How to Check if Your Clinic Appears in ChatGPT (and Track It)
A step-by-step way to find out whether ChatGPT names your clinic when patients ask for a recommendation — what to type, why answers change between runs, and how to turn a one-off check into a trend you can track.
Ask like a patient, not like an owner
The most common mistake is typing your own clinic name. That tests recognition, not recommendation. A patient rarely knows your name yet — they type “recommend a dermatology clinic near Gangnam station” or “my tooth aches, where should I go in Dongtan”. Those are the questions that decide whether you get the visit.
One check is an anecdote
Generative answers move. Ask the same thing twice within an hour and the list of clinics can differ. Engines also shift when the underlying search results shift — a change on the search side moves every clinic at once, which is easy to mistake for something you did. The fix is boring but decisive: freeze the question set and repeat it on a schedule.
Track competitors in the same sheet
Recording only your own appearances tells you very little. If your exposure drops from 40% to 25%, the useful question is whether the clinics around you dropped too. Logging which competitors were named in each answer turns a number into a diagnosis.
Do not blend the engines
ChatGPT and Perplexity mostly ground on Bing’s index; Gemini on Google’s; Claude cites a narrower set of sources. A clinic can sit at 90% in one engine and 0% in another, and a single averaged figure hides exactly the gap worth fixing. Keep the columns separate.
FAQ
How do I check if my clinic appears in ChatGPT?
Ask the way a patient would, not by your clinic name. Use a neighbourhood plus a specialty or symptom — for example “recommend a dentist in Dongtan” or “where should I go in Gangnam for implants”. Searching your own clinic name only tells you whether the model knows the name, not whether it recommends you.
Why does the answer change every time I ask?
Generative engines sample from a distribution and, when web grounding is on, read a live set of search results that also shifts. The same question can name different clinics minutes apart. That is why one check is an anecdote and a fixed question set repeated on a schedule is a measurement.
Does the app give the same answer as the API?
Not always. Web search may be on or off, and account settings and memory can influence the reply. Any tracking that claims to reflect what patients see should keep web grounding on and disclose that gap rather than ignore it.
How often should I track it?
Weekly is mostly noise for this kind of data. A two-week cadence over three months gives six points — enough to see a trend line rather than a single snapshot.
What should I record each time?
The exact question, the engine, whether your clinic was named, its position in the list, and which competitors appeared. Without the competitor column you cannot tell whether you slipped or the whole market moved.
Boily runs this as a service for clinics in Korea — a frozen question set across ChatGPT, Claude, Gemini and Perplexity, measured every two weeks, with competitor comparison in the same report. We do not guarantee rankings or exposure.