AI share of voice: how much of AI’s recommendations go to your brand
When someone asks AI for brands in your category, the answer names only a few. Share of voice measures how many of those mentions are yours compared with your competitors, engine by engine.
Short answer
AI share of voice is your percentage of all the brand mentions AI engines make when they answer questions in your category that don’t name your brand. Refractia calculates it in ChatGPT, Claude, Gemini, Meta AI and Google AI Overviews, pairs it with the discovery rate (how many of those questions you appear in) and shows every figure with its sample, as k of n. Chat engines and AI search are reported separately.
What AI share of voice is
It’s the AI-answer version of a classic marketing metric: your share of the conversation against your competitors’ share. What changes is where the conversation happens. When someone asks “Which digital bank is best in the United States?”, the engine doesn’t return ten links: it names a few brands and explains why. Share of voice measures how many of those mentions are yours.
Only organic questions count, the ones that don’t include your name. If the question is “What do you think of Brand X?”, the answer names Brand X every time, and counting it would inflate the number without saying anything about whether AI recommends you. What matters is whether you show up when nobody was looking for you.
How it’s calculated
- 01
Organic questions
From the Prompts Library, Refractia takes the category, comparison and purchase-intent questions that don’t name your brand. All of them are in your market’s language and name the country.
- 02
Mentions per brand
In each answer, it counts the brands named: yours and those of the competitors you declared.
- 03
Share of voice
Your organic mentions divided by the total organic mentions of all brands. If a scan has 200 mentions and 30 are yours, your share of voice is 15%, shown as 30 of 200.
- 04
Discovery rate
The percentage of organic questions in which you appear at least once. It measures reach: in how many questions AI remembers you, not how much space it gives you in each one.
- 05
Per engine, and ranked
The calculation is repeated for each engine and ranked in the competitive arena, which orders your brand and your rivals by organic citations, discovery rate and visibility over a 30-day window.
Why every figure comes with its sample
A 25% share of voice can be 1 of 4 mentions or 250 of 1,000. The percentage is the same, but the first figure can’t support a decision and the second can. That’s why Refractia always shows the numerator and denominator (k of n) next to each rate.
The other rule is that not measured is not zero. If an engine didn’t answer a question, that question drops out of that engine’s calculation. Counting it as “answered and didn’t name you” would lower your share of voice because of an engine failure, not because of anything about your brand.
Chat and AI search, reported separately
Refractia reports the chat engines (ChatGPT, Claude, Gemini and Meta AI), which answer from what the model knows, separately from Google AI Overviews, which summarizes what ranks on the web for that search in your country. A brand can have a strong share of voice in one type and almost none in the other.
Blending them gives a misleading average. A new brand with good web content can show up in Google AI Overviews because the engine paraphrases its pages, while staying unknown to the chat models. An established brand can be very present in the models and lose in search to whoever publishes better. They’re different problems, and you work on them differently:
| Pattern | What it usually means | Where to look |
|---|---|---|
| High share in search, low in chat | AI finds you on the web, but the model doesn’t have you in mind. | Influence Sources: who cites you outside your own site. |
| High share in chat, low in search | The models know you, but other pages rank better for those searches. | GEO Optimization: how AI reads your pages. |
| Low discovery rate | AI doesn’t associate you with the category, or with part of it. | Performance by Topic: which topics you disappear from. |
| High discovery, low share | You appear in many questions, but rivals get named more often. | AI Perception: how you’re described next to them. |
How to use the result
- Compare against your rivals and your previous scan, not against an ideal number. There’s no universal “good” share of voice: it depends on how many brands compete in your category.
- Look at each engine before the total. A big gap between engines tells you where the problem is.
- Break the average into topics: Performance by Topic shows which parts of the conversation you win and which you’re missing from.
- Follow the trend in Evolution, which compares share of voice across scans for your brand and each competitor.
For more on the metric itself, see the AI share of voice guide.
Frequently asked questions
How do you measure share of voice in ChatGPT?
Ask ChatGPT many questions in your category that don’t name your brand, count the brands it names in each answer, and calculate what share of those mentions is yours. Refractia does the same in four other engines and shows each result as k of n.
What is a good AI share of voice?
There’s no universal number: it depends on how many brands compete in your category. The useful comparison is against your rivals in the same scan and against your own previous scan.
Why does my share of voice vary so much between AI engines?
Because each engine answers differently. Chat engines use what the model knows and Google AI Overviews summarizes what ranks on the web, so a brand can be strong in one and absent from another.
What’s the difference between share of voice and discovery rate?
Share of voice measures what share of all mentions is yours. Discovery rate measures what percentage of the questions you appear in at least once.
How often is AI share of voice updated?
With each scan, which a person on your team or at Refractia starts. It isn’t continuous monitoring: you see change by comparing scans in Evolution.
Keep reading
See who AI names in your category
Your customers already ask AI for recommendations; today you don’t know who it names. We’ll show you what ChatGPT, Claude, Gemini, Google and Meta AI say about your brand and your competitors. Free, no commitment, and we reply within one business day.