Share of voice in AI: how to calculate your brand's share of answers

What is share of voice in the context of AI?

Share of voice in AI answers is your brand's share of the answers that models such as ChatGPT, Gemini, Perplexity or Google AI (AI Overviews) generate for specific user questions. What counts is whether the brand is mentioned and how often that happens compared with other brands in the same category.

You can measure this share in two ways. In the classic variant, you divide the number of mentions or recommendations of a given brand by the total number of mentions of all analyzed brands and multiply by 100%. In the Mention Share variant, you divide the number of answers that include the brand by the number of all answers, again multiplying by 100%. The first shows your slice of the pie relative to competitors; the second shows how often you appear in answers at all.

In AI, what matters most is whether the model mentions the brand in a specific answer to a specific question. You are not counting position within an answer, only the presence of a mention. That is why the whole methodology rests on asking consistent questions (prompts) repeatedly, collecting results consistently and comparing them for your brand and a group of competitors.

Why measure share of voice in AI?

First of all, user behavior is changing. In 2024, Gartner predicted that by 2026 the volume of queries in traditional search engines would fall by 25% in favor of AI chatbots and other virtual agents. If your brand is missing from AI answers, you lose part of the visibility that used to come from search engines, and adapting your content to user questions in a systematic way helps you keep and win it back.

AI is also increasingly influencing purchase decisions. According to a 2025 study by Capgemini Research Institute, 58% of consumers have replaced traditional search engines with generative AI tools when looking for product and service recommendations. If a model takes part in purchase choices, your brand's share of answers becomes a new KPI. It is a day-to-day metric that tells you whether your offer is being considered when a user asks a model for a recommendation or advice.

How to measure share of voice in AI answers

Define the category and competitors

Start by defining precisely the category in which you want to be mentioned. It should match how users ask, the way they would talk to an advisor rather than a search engine. Add a list of comparison brands: direct competitors and brands that AI may naturally associate with your use case. If your brand can be written in several ways, include those variants so that mentions are recognized correctly.

Next, set counting rules, meaning what you treat as a mention: the full name, an abbreviation, the domain name, an umbrella product name. Stick to them throughout the process, because consistency is the condition for comparable results.

Build a prompt basket

Base the basket on user intents. Prepare problem-based, comparison and recommendation questions that could realistically come up in a conversation with AI, in natural language and in different phrasings. Cover the whole decision funnel, from educational questions through choosing a solution and purchase criteria to implementation and after-sales support, and you will see at which points in the conversation AI mentions you most often and at which least often.

Also add context: industry, company size, market or constraints, exactly as a real user would. Context changes the answer, and therefore the share of voice you observe for your segment. For each prompt, assign a clear evaluation criterion: whether the brand was mentioned (yes/no), whether it is a recommendation or a neutral mention, and which sources the AI cites.

Test across different AI models

Models are not uniform: they differ in how they answer and in the sources on which they base recommendations. Testing across several engines reduces the influence of any single tool's quirks. Run the measurement itself in a fixed order:

1. Run the prompt basket in several AI engines, always with the same questions, in the same language and context.
2. Save the full answers so you can judge whether a mention meets your counting rule.
3. Calculate Mention Share: number of answers with the brand / number of all answers × 100%.
4. To compare brands in the category, calculate share of voice: number of brand mentions / total number of mentions of all brands × 100%.
5. Note the blind spots, meaning questions where AI mentions competitors and not your brand.

Blind spots are the best input for a content and PR action plan.

Tools for measuring share of voice in AI

The simplest tool is a spreadsheet in which you map prompts across columns: AI model, prompt, whether the brand is mentioned, mention type, co-occurring brands, cited sources and notes. It gives you full control over the methodology and is easy to extend with new questions. It is worth supplementing it with a register of the domains and publications that AI refers to, which is a practical hint about where to strengthen your expert presence or fill gaps in your content.

When you need automation, reach for a system for regular monitoring of AI visibility. Attent monitors brand mentions in four engines (ChatGPT, Gemini, Perplexity and Google AI) and shows share of voice against competitors, blind spots, sources cited by AI, alerts about changes and a ready-made leadership report.

Regular measurement of share of voice in AI answers is not a one-off audit but a standing routine for managing demand. Without comparable measurements over time, it is hard to tell noise from trend and to plan actions that will really increase the brand's share of conversations with AI.

Analyzing the context of AI answers

The percentage of mentions alone is not enough for sound decisions. First, distinguish neutral presence from a recommendation: both mean visibility, but only a recommendation shows that the model actively suggests the brand to the user. Also check what the AI gives as the reason for your presence (expertise, offer, price, reviews or sources), because that hints at what to strengthen in your content and reputation.

Co-occurring brands say a lot. Who you are mentioned alongside defines the shelf on which AI places you. If you consistently land in a different category than the one you are fighting for, you need to correct your messaging and expert visibility. When models base their conclusions on specific domains, publications or reports, check whether your content is present there and detailed enough.

Look at the trend, not a single reading. Differences between successive measurements may result from model or source updates, so only several measurements show which way the brand's share is heading.

Optimizing content to increase share of voice

If AI does not mention the brand for problem-based or comparison questions, prepare materials that solve those queries directly, with a clear structure, definitions and examples of use. For the questions where you are missing, create dedicated pages or sections that answer the specific need, and use PR to strengthen credibility in the category.

Since models refer to particular sources, make sure your materials are present in those ecosystems: through expert publications, guest articles or by expanding your own content so that it is cited more readily. Also make it easier for models to recognize the brand by using names and spelling variants consistently on your site and in your materials.

After each change, check in the next measurements whether the share of mentions is growing. It is a repeatable cycle: diagnosis, content, verification, correction.

Examples of applying share of voice measurement

Three implementation scenarios show how to use the measurement in practice, without attributing results to real brands.

In a product scenario, the team defines the category and a list of alternatives, creates a basket of comparison and recommendation questions and runs the test across several models. The analysis reveals that the brand is not mentioned for questions about implementation and after-sales support. The team prepares a series of guides and expert materials, then repeats the measurement to check whether mentions have appeared.

In a geographic scenario, the team studies visibility at the national and local level, and the prompt basket includes variants with regional context and industry specifics. The first measurement produces a map of blind spots for local queries, from which a content and PR plan is built, aimed at the media and sources that models readily cite.

In an assortment scenario, each product line gets a separate prompt basket. The analysis shows on which topics the brand is mentioned together with particular competitors, and the team tidies up its messaging to avoid segments being confused and to anchor itself better on the target shelf.

All three rest on the same axis: a fixed prompt basket, tests across multiple models, a measurable share of mentions, and work on blind spots and sources cited by AI.

Summary and conclusions

Measuring share of voice in AI answers is an ongoing process: you define the category and competitors, build a prompt basket, test across several models, calculate the share of mentions and analyze the context of the answers along with their sources. You calculate the indicators with simple formulas: Mention Share, which is number of answers with the brand / number of all answers × 100%, and classic SOV, which is number of brand mentions / total number of mentions × 100%. You turn the findings into content and PR actions, especially where you see blind spots.

If you want to do this regularly and in a comparable way, you can use a monitoring tool for AI answers. Attent automates asking questions in four engines, tracks share of voice against competitors, catches blind spots and points to the sources cited by AI, which makes it easier to plan specific actions. You can book a call about monitoring at getattent.com.

Sources

1. Gartner – Gartner Predicts Search Engine Volume Will Drop 25% by 2026, Due to AI Chatbots and Other Virtual Agents, 2024
2. Capgemini Research Institute – What matters to today's consumer 2025 (komunikat prasowy), 2025