What Is AI Share of Voice?

AI share of voice is the percentage of all provider mentions, across a fixed set of AI answers, that belong to your practice. For a health practice, it shows how you stack up against local competitors inside the answers patients actually read. If AI tools named 100 providers across your test runs and 18 were you, your share of voice is 18 percent.

How AI share of voice works

The idea comes from advertising and search, where share of voice means your portion of all the attention in a category. In AI search, the category is a set of patient questions, and the attention is the list of names in the answers.

Patients get a short list. An AI answer is a written summary, not a page of ten results, so each name in it takes up one of a few slots. Share of voice shows who is taking those slots. This is the main reason it is different from AI visibility, which asks only how often you appeared.

The calculation has four steps.

• Fix the list of patient questions, covering your services and cities.
• Run each question several times on each tool.
• In every answer, write down every provider named, including you.
• Add up all mentions for the total, then divide your mentions by the total.

Here is why the denominator matters. Suppose a med spa is named in half of its test runs, which sounds good. But three competitors are each named in nearly every run. The med spa has a fair visibility score and a small share of the conversation. Only the second number tells it how it stacks up.

Share of voice is only meaningful if the method is steady. The same questions, the same tools and many runs are required, since answers vary from run to run. A single run gives an arbitrary number. For ChatGPT checks, use a temporary chat, which OpenAI describes as starting without memory by default, so your own history does not tilt the answer.

Report each tool separately. Patients use different tools, and a single blended number hides which one is helping or hurting you. Keep a line for each tool in your sheet.

Here is why a relative measure is worth the extra work. Your own visibility can hold steady while your share falls, simply because a competitor started appearing more often. Without the total, you would not see it happen.

A strong share in one tool can hide a near-zero share in another.

Counting presence alone leaves out two things. One is where in the answer you appear, since a name listed fifth in every answer is likely weaker than a name listed first. The other is how crowded the answer is, because one of three names and one of thirty are very different positions. If you want a score that reflects either, you must invent your own weights, and I would say so plainly in any report. They are a judgment call, not a standard. Also note which practices keep showing up together in the same answers. That group is your real competitive set, and it is often not the practices you expected.

Share of voice counts names, so it relies on AI brand mentions. Count a name when it appears in the answer text, whether or not a link comes with it.

If you also want to know how often your website was linked, that is the related measure citation share. Keep the two scoreboards apart, because a practice can lead one and trail the other.

You can also slice the number. Compute it per service, such as hormone therapy versus weight loss, or per city. The overall figure hides the places where you are strong and the places where you are absent.

Example: AI share of voice at a health practice

Take a med spa in a mid-size city. It builds a list of 20 questions about injectables, laser treatments and skin care, and runs each one three times in ChatGPT. That is 60 runs. This example is illustrative, not a transcript, and the figures are made up to show the arithmetic.

Across the 60 runs, the tool names providers 240 times in total. The med spa is named in 36 of those runs, one mention each.

• Total provider mentions, 240.
• The med spa's mentions, 36.
• Share of voice, 36 divided by 240, which is 15 percent.

Now break it down by service. In injectables, the med spa holds 8 of 120 mentions, about 7 percent. In skin care, it holds 28 of 120, about 23 percent. The owner now knows where to put effort. Injectables questions are dominated by other names, and skin care is a position to defend.

One more point. The med spa was named in 36 of 60 runs, a visibility of 60 percent. Yet its share of voice is 15 percent, because every answer names several providers and the med spa takes only a slice. The two numbers answer different questions.

For real data on how often practices get named at all, here is the Health Biz Scale test. We checked 764 established US health practices between May 11 and Aug 20, 2026, with one check per practice and test runs excluded. We asked GPT-4o-mini, through OpenAI's API and answering from memory with no web search, 2,292 patient-style questions. Only 6 of the 764 practices (0.8%) were named. The ChatGPT app searches live and can answer differently, so that is a memory-only baseline and not what patients see in the app. It suggests that a typical established practice may hold a small share, which is why measuring it is the first step.

How to check yours in 60 seconds

1. Ask one patient question three times, each in a separate ChatGPT temporary chat.
2. List every provider named in each answer, including yours.
3. Count your mentions and the total mentions.
4. Divide yours by the total. That is a rough share for that question.

Three runs of one question is a rough start. It shows you the arithmetic and the competitors, and it is not a score you should report.

Write down the names that keep showing up. Those are the practices holding the slots you want, and you will study them later.

What to do about it

1. Build a fixed list of 20 to 30 patient questions across your services and locations.
2. Run each one several times on the same AI tools, on the same schedule.
3. Record every practice named in every answer, in a spreadsheet.
4. Add up all mentions, then divide your mentions by the total for your share.
5. Compute it per question group, such as hormone therapy or weight loss, to see where you are strong. See prompt tracking for how to keep the list steady.
6. Review monthly and compare against the same months, not single days.
7. Keep the questions identical from one month to the next. If you must add a question, add it to a new list and keep the old one running so the trend stays clean.
8. Study the practices that hold the most mentions. Look at their pages, profiles and press, and close the gaps with your own. Noting which outside sites those answers lean on gives you an AI source map to work from.