How to audit your clinic's AI visibility in an afternoon: twenty prompts, three engines, one score

Choose the prompts

Write twenty questions in the words a patient would use, not the words a marketer would. Eight discovery prompts ("best facelift surgeon in [city]", "where should I go for All-on-4 in [city]"). Six comparison prompts ("deep plane vs SMAS, who to see in [city]", "is Istanbul safe for a hair transplant or should I stay in the UK"). Three price prompts ("blepharoplasty cost [city]", "how much is a single implant in [city]"). Three by-name prompts ("is [clinic] reputable", "what do patients say about [clinician]", "[clinic] reviews"). Use your real procedures and your real city.

Run them

ChatGPT, Gemini and Perplexity, logged out or in a fresh session, each prompt run three times across different days. Answers change run to run, so a single screenshot proves nothing either way. Record, for each run, the clinics named, whether yours was recommended or merely mentioned, and the sources the engine cited when it showed them.

Score it

Three states per cell: named and recommended; mentioned but not chosen; absent. Count the share of runs in each. For the free Business Scan, the share of discovery and comparison runs where the practice is named is the core of the AI visibility point; the by-name prompts tell you what the engines believe about you, which is a separate finding and often the more alarming one (one engine credited a client's identity to another operator entirely).

Read the sources

This is the part most people skip and the part that tells you what to do. Write down every source the engines cited: review platforms, directories, forums, listicles, videos, clinic websites. Whitespark's July 2026 study found 77 percent of citations for local queries are off-page. Your list will show you which pages are carrying the answer in your city, and therefore where your practice needs to be present.

What to do with it

If you are absent from discovery prompts, the work is visibility: credentials, procedure pages, Google profile, reviews, presence where the engines read. If you are mentioned but not chosen, the work is reputation: review velocity, responses, the specific procedures reviews mention. If the by-name prompts return errors, the work is entity consistency: the same name, address, clinicians and facts everywhere you appear. Repeat the panel monthly and watch the trend, not the position.

What this does not tell you

How many patients arrived from an AI answer. That needs attribution in your booking system, which is point ten of the audit. It also does not tell you why the engine chose the clinics it did, only where it looked. That is enough to act on.