What is share of voice in AI search for clinics?

It is the proportion of all clinic mentions across a round of questions that belong to one clinic. Where appearance rate answers 'did we show up', share of voice answers 'who is occupying the answers in this area'. It is calculated per engine over a frozen, unbranded question set, after normalising name variants so one clinic is not counted as several.

How is it different from appearance rate?

Appearance rate is about you alone: the share of questions where your clinic was named. Share of voice is about the category: of every clinic named across those answers, how many mentions were yours. Both can move independently. Your appearance rate can rise while your share falls, because more clinics in the area are also being named.

Can I use share of voice in my clinic's advertising?

In Korea this is risky. Advertising that compares a medical institution with other medical institutions is restricted under the Medical Service Act, and a chart showing your clinic against named competitors would be exactly that. Share of voice is useful as internal analysis to decide what to work on. It should not become patient-facing marketing material. This is general information, not legal advice.

Why do competitor names come out messy?

Engines write clinic names inside prose, so raw extraction picks up sentence fragments, generic words, and several spellings of the same clinic — full name, short name, branch name, district prefix. These have to be merged before counting. In one of our Korean dental measurements the raw list of 193 apparent competitors reduced to 70 real ones after normalisation.

Does a high share of voice mean more patients?

Not directly. A mention is not a link, and a link is not a visit. Some answers name a clinic without citing its site, in which case no traffic follows. Tracking whether AI answers produce visits is a separate measurement based on referrer data, and it undercounts because AI apps often send no referrer at all.

SHARE OF VOICE · AI SEARCH

Share of Voice in AI Search — Which Clinics Occupy the Answers

2026-09-14 · Boily

Appearance rate asks whether you showed up. Share of voice asks who is taking the slots. Across a frozen set of unbranded patient questions, count every clinic named in every answer, per engine, after merging name variants. Your share is your mentions over all mentions. It tells you whether a low appearance rate means the category is crowded, empty, or not answering with clinic names at all — three situations that call for different work. Boily measures this for Korean clinics every two weeks and does not guarantee visibility.

How it is calculated

Count mentions, not answers

For every question in the frozen set, on every engine, list each clinic named in the answer. One answer can name three clinics, so the denominator is total mentions across the round, not the number of questions. A clinic's share of voice is its mentions divided by all mentions in that round.

Normalise clinic names first

The same clinic appears as a full legal name, a short name, a branch name, and sometimes with the district attached. Without merging these variants the count fragments and every clinic looks smaller than it is. In one Korean dental measurement, raw name extraction produced 193 apparent competitors that collapsed to 70 after merging variants and removing sentence fragments.

Keep it per engine

A clinic can hold a large share on one engine and none on another, because the engines read different sources. A blended share of voice averages away the only actionable information.

Separate branded from unbranded questions

Questions containing the clinic's own name will almost always return that clinic. Including them inflates share of voice for everyone and makes rounds incomparable when the question mix changes. Report unbranded questions as the headline and branded ones separately as a recognition check.

Reading it against your own rate

The pair of numbers is more informative than either alone. Four common patterns and what each implies.

Low appearance, concentrated share elsewhere: A small number of clinics occupy most mentions. The slots are taken and the sources behind those clinics are worth examining — what is cited about them, and what kind of document it is.
Low appearance, flat share across many clinics: No one owns the category. Answers name a scattered set, which usually means the engines are assembling from thin data. This is the condition where a clinic with complete, readable facts tends to move fastest.
Low appearance, few mentions of anyone: The engine is not answering with clinic names at all for these questions — it may be returning general guidance or map-style entries. The fix is a different question set, not more content.
Rising appearance, falling share: You appear more often but so does everyone. The category is filling up. This is normal as more clinics in an area publish structured information, and it is the reason a single absolute number can look flat while position improves.

Five traps

· Comparing share of voice across rounds with a changed question set. Any edit to the questions breaks the comparison.
· Treating a competitor's high share as proof that a specific tactic works. You can see what was cited, not why the engine chose it.
· Reading a single round as a trend. Mention counts move run to run even with identical inputs.
· Publishing competitor comparisons in clinic marketing. In Korea, advertising that compares a clinic with other medical institutions is restricted under the Medical Service Act, so share of voice belongs in internal analysis, not in patient-facing material.
· Assuming a mention is a link. A clinic can be named with no citation to its own site, in which case no traffic follows the mention.

What replaces rank for a single clinic is covered in AI search ranking for hospitals, and what to look for in a measurement tool is in AI search visibility tools for clinics.

Frequently asked questions

Q. What is share of voice in AI search for clinics?: A. It is the proportion of all clinic mentions across a round of questions that belong to one clinic. Where appearance rate answers 'did we show up', share of voice answers 'who is occupying the answers in this area'. It is calculated per engine over a frozen, unbranded question set, after normalising name variants so one clinic is not counted as several.
Q. How is it different from appearance rate?: A. Appearance rate is about you alone: the share of questions where your clinic was named. Share of voice is about the category: of every clinic named across those answers, how many mentions were yours. Both can move independently. Your appearance rate can rise while your share falls, because more clinics in the area are also being named.
Q. Can I use share of voice in my clinic's advertising?: A. In Korea this is risky. Advertising that compares a medical institution with other medical institutions is restricted under the Medical Service Act, and a chart showing your clinic against named competitors would be exactly that. Share of voice is useful as internal analysis to decide what to work on. It should not become patient-facing marketing material. This is general information, not legal advice.
Q. Why do competitor names come out messy?: A. Engines write clinic names inside prose, so raw extraction picks up sentence fragments, generic words, and several spellings of the same clinic — full name, short name, branch name, district prefix. These have to be merged before counting. In one of our Korean dental measurements the raw list of 193 apparent competitors reduced to 70 real ones after normalisation.
Q. Does a high share of voice mean more patients?: A. Not directly. A mention is not a link, and a link is not a visit. Some answers name a clinic without citing its site, in which case no traffic follows. Tracking whether AI answers produce visits is a separate measurement based on referrer data, and it undercounts because AI apps often send no referrer at all.

Your rate and the category, in one report

Boily measures four engines against a frozen question set every two weeks for Korean hospitals and clinics, reporting appearance rate, engine split and share of voice with competitor names kept to internal analysis. We do not guarantee visibility or position.

About Boily Our own numbers

Written by Boily. Observations come from our own biweekly measurement of Korean clinics; samples are small, clinic names are withheld, and no causal claim is made. The note on Korean medical advertising rules is general information, not legal advice. Boily is a vendor in this market and does not guarantee visibility or position.