AI Share of Voice

AI share of voice is the AI-era adaptation of traditional share-of-voice measurement. For a defined set of category queries, it measures the percentage of brand mentions in AI-generated responses that are yours, relative to competitors. If ChatGPT mentions three brands when answering "best running shoes for marathon training" and yours is one of them, you have roughly 33% share of voice on that query on that platform.

It is the most direct competitive measure in AI search. Where found rate tracks whether you appear at all, share of voice tracks how much of the answer is yours relative to the field. A brand can be cited every time on its own name and almost never on its category. Both matter, and they need different work.

HubSpot defines share of voice in the answer-engine context as how often and how prominently AI platforms mention your brand in category conversations relative to competitors. The prominence dimension is meaningful: being mentioned third in a list of five is worth less than being mentioned first, and being named in the body of the answer is worth more than only being cited in a footnote.

A consistent finding in AI search measurement: the same brand, with the same content, measured on the same questions, can have very different share of voice and sentiment on different assistants. Superlines' early 2026 analysis of a single brand found a large sentiment gap between Perplexity and ChatGPT on identical questions.

The reasons are structural:

Different data. ChatGPT combines training data with live browsing. Perplexity retrieves from a live index on every question. Gemini leans on Google's index and Knowledge Graph. A brand strong on review sites tends to do well on Perplexity. A brand strong on Reddit and community content tends to do well on ChatGPT.

Different retrieval. Perplexity favours fresh, citable sources. ChatGPT's browsing is more selective. Google AI Overviews draw on the regular index with their own signals.

Different style. Gemini tends toward balanced multi-brand comparisons, ChatGPT toward a single confident recommendation, Perplexity toward structured pros and cons. The same question produces a different share on each.

So measure per assistant. A brand averaging 30 percent might be at 50 on Perplexity and 10 on ChatGPT. The average hides the gap, and the gap is the backlog.

The pattern that works:

1. Define the question set. 30 to 100 category questions your shoppers would actually ask ("best [category] for [use case]", "[category] under [price]", "alternatives to [competitor]"). Keep it fixed so results compare over time.
2. Run it on a schedule across assistants. ChatGPT, Perplexity, Google AI Overviews, Google AI Mode and Gemini. Record the full answers.
3. Extract mentions. Count brand and product mentions, yours and competitors', with position (first, middle, last) and context (recommended, considered, mentioned in passing).
4. Compute share. Per assistant per question, then across the set. Watch the trend.
5. Weight prominence. A mention in the first sentence is worth several at the end.

What improves it is what improves AI visibility generally: answer-first content, complete schema, third-party reviews, entity clarity, freshness and, for retailers above all, product information that answers the category question so the assistant can show a product card. The Paz platform runs this measurement for you, and a Paz expert turns the low-share assistants and questions into product information fixes, shipped within your guidelines. See the AI visibility service.