How to measure your AI Share of Voice: the KPI traditional SEO can't see
Google Analytics tells you how many people arrived at your site. It doesn't tell you how many people didn't arrive because an AI sent them to a competitor. That's what AI Share of Voice measures.
AI Share of Voice = the number of times your brand appears mentioned in AI engine responses, measured across a standardized set of queries relevant to your sector. Expressed as X mentions out of Y total queries (e.g., 14/40).
Why analytics dashboards don't see the problem
When someone uses Google and clicks your site, Google Analytics records that session with source "google / organic." When someone uses ChatGPT and your brand appears recommended, that's not recorded in any dashboard. If the user clicks the link ChatGPT cites, Analytics sees it as "direct" — no referrer. And if the user simply notes your name and searches for you later, there's no way to trace the origin back to AI.
The result: the impact of AI engines on your business is invisible to traditional metrics. You can't see how many customers are discovering you through ChatGPT, nor how many potential customers AI is sending to your competitor instead of you.
AI Share of Voice exists to make that layer visible. It doesn't measure the traffic that arrives — it measures the presence your brand has in the channel where decisions are being made.
How to measure it: the 5-platform, 40-query protocol
Queries must represent how a real customer would search for what your company offers. For a marketing consultancy: "best digital marketing agency in [city]", "which GEO agency do you recommend", "how to improve my ChatGPT visibility." For an e-commerce: "where to buy [category] online", "which is the best [product] store online." The 40 queries are distributed across 4 categories: brand queries, category queries, problem queries, and comparison queries.
ChatGPT, Perplexity, Gemini, Claude, and Copilot. Each query runs in "new conversation" mode to eliminate the contextual memory effect. Responses are saved in full.
For each response, record whether the brand is mentioned, at what position (first mention, second, etc.), the context of the mention (positive, neutral, negative), and whether a link is cited. This produces a table of 200 rows (40 queries × 5 platforms).
Total Share of Voice = total mentions / total queries. Also calculated per platform (Share of Voice in ChatGPT = X/40) and per query category. The result is a number that can be tracked over time.
The same protocol is run for the 3 to 5 main competitors. This turns your own Share of Voice into a relative number — knowing you have 12/40 is useful; knowing competitor A has 28/40 and competitor B has 6/40 is what defines the strategy.
How to interpret the results
| AI Share of Voice | Diagnosis | Priority |
|---|---|---|
| 0–5 / 40 | AI ignores the brand or confuses it with others | Urgent: technical foundation + foundational content |
| 6–12 / 40 | Appears in direct brand queries but not category queries | High: expand to problem and comparison queries |
| 13–20 / 40 | Solid category presence, with gaps by platform | Medium: close the platform-by-platform gaps |
| 21–30 / 40 | Category leadership | Defensive: maintain and monitor competitors |
| 30+ / 40 | AI dominance for that category | Expand to new query categories |
Why results vary between platforms
Each AI engine has a different search infrastructure: ChatGPT (OpenAI) uses Bing for real-time search. Perplexity AI primarily uses Google. Gemini uses Google Search directly. Claude (Anthropic) uses Google in real time. Copilot (Microsoft) uses Bing exclusively.
This means if your company ranks well in Google but not Bing, you'll have better Share of Voice in Perplexity, Gemini, and Claude than in ChatGPT and Copilot. Measuring all 5 separately lets you identify exactly where the gap is and prioritize accordingly.
In the audits we run for companies in Argentina and Latin America, 78% start with a Share of Voice between 0 and 4 out of 40 queries. 91% show significant differences between platforms — typically better in Gemini and Perplexity than in ChatGPT and Copilot. That's almost always explained by better Google ranking than Bing ranking.
Frequency and methodology consistency
Measure AI Share of Voice once a month. Models don't change their perception of a brand week to week — changes are gradual and consolidate over weeks or months. The critical thing: use the same 40-query set each measurement. If you change the queries each month, you can't tell whether a change in the number is due to your GEO work or to asking differently.
FAQ
What is a good AI Share of Voice?
As a reference: 0/40 is the typical starting point. 10/40 to 20/40 is a solid result at 90 days. 25/40 or more is category leadership. In sectors with little GEO competition, 15/40 is reachable with relatively simple interventions.
How often should I measure?
Once a month is enough. The critical thing is always using the same 40-query set so the data is comparable.
Why is my result different in ChatGPT vs Perplexity?
Different search infrastructure: ChatGPT and Copilot use Bing; Perplexity, Gemini, and Claude use Google. Better Google ranking than Bing ranking explains most of these gaps.
What is your AI Share of Voice today?
We measure it with 40 standardized queries across 5 platforms. In 7 days you know exactly where you stand and what to do to improve.
Request GEO audit →