What counts as being named?
A brand counts as named when the answer presents it as a relevant option, example or source for the question asked. We do not count:
- A passing mention in an unrelated context.
- An appearance caused only by the question naming the brand. Branded and unbranded questions are reported separately.
- A mention that misidentifies the brand. That is logged as an accuracy issue instead.
Unbranded questions matter most. They are the questions buyers ask before they know your name, which is where shortlists are formed.
How is Answer Share calculated?
An eligible question is one where a brand like yours could reasonably appear: within your offer, your market and your audience. Because AI answers vary from one run to the next, every question is asked several times on each platform and the results are averaged. A question where you appear in two of three runs contributes 0.67, not 1.
Answer Share = questions in which you are meaningfully named ÷ eligible questions tested
How is it broken down?
By platform
ChatGPT, Gemini, Perplexity and Copilot often name different brands for the same question.
By buying stage
Brands are often visible when buyers explore, then disappear when they compare.
By question type
Recommendations, comparisons, explanations and validation checks each behave differently.
Against competitors
Who is named when you are not, and how often.
What sits alongside Answer Share?
Answer Share tells you whether you are in the answer. Five companion measures tell you how:
Recommendation Share
How often you are presented as an option worth considering, not just mentioned.
Shortlist Share
How often you make the finite list of options in recommendation answers.
Citation Share
How often the answer draws on your own content or third-party sources about you.
Accuracy
Whether what AI says about you is correct, current and supported.
Narrative
The words AI repeatedly uses to describe you.
What Answer Share is not
It is not website traffic, it is not a search ranking, and it is not a sales forecast. It measures whether you are considered. We also do not roll it into a single score out of 100: a number that blends visibility, accuracy and sentiment is easy to sell and hard to defend.
Illustrative example · fictional brands
Harbourline Hotels
We test 80 unbranded questions a leisure traveller might ask, on four platforms, three runs each: 960 answers in total. Harbourline is named, on average, in 25.6 of the 80 questions, an Answer Share of 32%. The breakdown is more useful than the headline: 45% at the exploration stage, 18% at comparison. Harbourline is being found, then losing the shortlist. That points the work at comparison content and third-party reviews, not at being discovered.
Frequently asked questions
Is Answer Share the same as share of voice?
No. Share of voice counts mentions or impressions across media. Answer Share counts presence across the specific questions buyers ask AI assistants, question by question.
How many questions do you test?
It depends on the purpose: 20 to 30 for a quick pulse, 75 to 150 for a category study, and 200 or more for a benchmark.
How often does Answer Share change?
AI answers shift as models and sources change. We track it on the same question set over time, so movement reflects change in the answers, not in what we asked.
Can a small brand have a higher Answer Share than a market leader?
Yes, particularly on specific, contextual questions where clear, well-evidenced information matters more than size.