How it works
AI Share of Voice measures one thing: how often AI answer engines name your clinic when patients ask about the treatments you offer. You do not measure it against a keyword list. You measure it against a prompt set, because patients ask AI questions in full sentences.
The method is simple and repeatable:
- Build a fixed prompt set of 30 to 100 questions real patients ask. Things like "best place for lip filler in Oakville", "is Morpheus8 worth it", or "who does natural-looking Botox near me".
- Name your competitor set. Usually the five to ten clinics you actually lose consults to, not every med spa in the city.
- Run the same prompts on the same engines on a schedule. ChatGPT, Google AI Overviews, Perplexity, Gemini, and Copilot all answer differently.
- Record every brand mentioned in each answer, and whether your site was cited as a source.
- Divide your mentions by the total brand mentions across the run. That percentage is your AI Share of Voice.
Two details make the number honest. Keep the prompt set frozen between runs, or you are measuring your own edits instead of the market. And run each prompt more than once, because AI answers vary between sessions. A single answer is an anecdote. Fifty answers across a fixed prompt set is a measurement.
Why it matters for aesthetic clinics
Aesthetics is a considered purchase. Patients research for weeks before they book, and more of that research now starts in an AI chat instead of a page of ten blue links. When the answer names three clinics and yours is not one of them, you never enter the consideration set. There is no impression to win back later and no ad to retarget. You were quietly invisible.
Rankings will not warn you about this. You can hold the top spot on Google for "Botox Toronto" and still be missing from the AI answer to the same question, because the engines read differently. They pull from reviews, directories, third-party roundups, and verifiable facts about your business, not just from your own pages.
Share of Voice also reframes the goal in a way clinic owners already think. Ranking is a solo measure. Share of Voice is competitive. If your mentions climb from 8 percent to 20 percent, another clinic lost that ground. That is the same math as market share in your postal code, applied to the place patients now ask their questions.
AI Share of Voice vs AI Visibility
These get used interchangeably and they are not the same thing.
| AI Visibility | AI Share of Voice | |
|---|---|---|
| Question it answers | Do AI engines mention us at all? | How often do they mention us versus rivals? |
| What it measures | Presence and citation rate | Your percentage of all brand mentions |
| It moves when | You get cited more often | You get cited more, or competitors get cited less |
| Best used for | Diagnosing whether you exist to AI | Tracking competitive position in a local market |
Both matter, in order. Visibility tells you whether you are in the game. Share of Voice tells you whether you are winning it.
The Ownerized take
Most clinics find their AI Share of Voice is close to zero and assume the fix is more content. It usually is not. AI engines cite clinics they can verify, which means consistent business facts, a real review profile, clear treatment pages, and mentions on the sources the models already trust. We treat Share of Voice as the scoreboard for the whole growth system rather than a marketing metric, and we measure it across the nine surfaces patients actually ask on before changing anything. Putting a number on that gap is the first thing the AI Growth System audit does.
Benchmarks
Use these as operating guidance for a defensible measurement, not as performance targets:
- Prompt set size: 30 to 100 questions, covering your top treatments, your city, and the objections patients raise before booking.
- Competitor set: 5 to 10 named clinics. Wider than that and the percentage stops meaning anything.
- Repetition: run each prompt at least three times per engine per cycle, because answers shift between sessions.
- Cadence: monthly at minimum. Model updates can move results overnight.
- Stability: freeze the prompt set. Change it and you have reset your baseline.
- Reading the number: judge direction across quarters, not precision in a single month. One weak answer is noise. Three months of decline is a problem.
- Starting point: expect a low number on the first run if no one has worked on this. That is normal, and it is the point of measuring.
Frequently asked questions
How do I measure AI Share of Voice for my clinic?
Start with a fixed set of 30 to 100 questions your patients actually ask, name five to ten real competitors, then run those prompts on ChatGPT, Google AI Overviews, and Perplexity each month. Record every clinic mentioned. Your share is your mentions divided by all brand mentions. Freeze the prompt set so runs stay comparable.
What is a good AI Share of Voice?
There is no universal benchmark, and anyone quoting one is guessing. AI Share of Voice is local and category-specific, so judge it against the clinics you lose consults to. If you are one of five real options in your area, being named in far fewer than a fifth of answers means you are losing ground.
Is AI Share of Voice the same as an SEO ranking?
No. Ranking measures your position on one results page. AI Share of Voice measures how often engines name you inside an answer, versus competitors. You can rank first on Google and still be absent from the AI answer for the same question, because the engines pull from reviews, directories, and third-party sources too.
How often does AI Share of Voice change?
Answers vary between sessions, so a single check tells you almost nothing. Model updates can also shift results overnight. Measure monthly, run each prompt several times, and read the trend across quarters. Treat one bad answer as noise and a three-month slide as a real problem. Set the cadence once and hold it.
Can I improve AI Share of Voice without writing more content?
Often yes. Engines cite clinics they can verify, so fixing inconsistent business details, thin treatment pages, and a weak review profile usually moves the number before new articles do. Getting mentioned on directories and roundups the models already trust helps as well. Content matters, but it is rarely the first fix.