AI Share of Voice: A New Metric for Brand Visibility in the Age of Artificial Intelligence

In an evolving digital landscape, marketers now face a new metric to navigate: AI Share of Voice (AI SOV). This metric aims to quantify a brand’s visibility within AI-generated responses compared to its competitors. While initially met with a degree of skepticism, AI SOV can offer valuable insights into a brand’s presence in AI-driven search and information retrieval, provided it is calculated and interpreted correctly.

AI Share of Voice is defined as the proportion of AI visibility a brand captures relative to tracked competitors across a defined set of prompts. A common method for calculating AI SOV involves dividing the number of responses that mention a specific brand by the total number of responses that mention any tracked brand, then multiplying by 100. For instance, if a brand appears in two out of three tracked ChatGPT responses concerning email marketing software, and its two competitors also appear in two responses each, all brands would have a seemingly equal 67% AI SOV.

However, this basic calculation fails to account for the varying popularity and reach of different prompts. A prompt like "best email marketing software," which is likely to be queried more frequently, carries significantly more weight than a more niche question. To address this, tools like Ahrefs’ Brand Radar employ a more sophisticated approach by weighting each prompt by its "AI-adjusted volume." This metric combines the monthly Google search volume of the keyword underlying the prompt with a platform ratio. This ratio reflects how much traffic each AI platform directs to websites compared to Google organic search, derived from Ahrefs’ traffic data. For example, a keyword with 20,000 monthly searches might translate to 2,300 estimated impressions on ChatGPT, given its current platform ratio of 0.115.

When a brand is mentioned or its website is cited within an AI response, the AI-adjusted volume of that prompt contributes to the brand’s "estimated impressions." The AI SOV is then calculated by dividing a brand’s total estimated impressions by the combined estimated impressions of all tracked brands appearing in the responses. Using the previous example, if a brand’s mentions stem from less frequently asked prompts, its AI SOV share could drop to 31%, even with the same number of mentions. It is crucial to understand that estimated impressions represent a modeled exposure, not a definitive count of viewers.

The AI SOV calculation also varies across different AI platforms due to distinct user behaviors and traffic patterns. Currently, ChatGPT has a ratio of 0.115, Gemini 0.011, and Perplexity 0.01. Consequently, a prompt with a 20,000 search volume generates 2,300 impressions on ChatGPT but only 220 on Gemini. AI SOV is typically measured per platform, and overall scores are weighted averages, giving more prominence to platforms with higher impression volumes. This can sometimes obscure weaknesses on less prominent platforms. Therefore, it is advisable to examine platform-specific scores before considering combined results.

Crucially, any AI SOV measurement is contingent on the defined scope, including the prompts tracked, the platforms analyzed, the competitors included, the geographic location, and the time period. Changing any of these parameters will alter the resulting percentage. This explains why different tools might report varying AI SOV figures for the same brand; the discrepancy often lies in the differing methodologies and scopes. Always clearly report the specifics of the measurement, such as the brands, aliases, competitors, market, location, platforms, dataset, prompt group, tracking frequency, and date range, to ensure clarity and prevent misinterpretation.

AI SOV proves most valuable when answering targeted questions, such as identifying a brand’s overall AI visibility, pinpointing its performance on specific AI platforms, understanding its competitive standing in particular niches, or tracking its presence in relation to specific keywords and topics. For instance, a brand with a 12% overall AI SOV for "email marketing software" might reveal a stronger presence in broad comparison prompts but a weaker performance in specific sub-topics like automation. This granular insight is far more actionable than a general percentage, highlighting areas for strategic content creation, comparison page development, and outreach efforts. However, it’s essential to remember that visibility does not equate to tangible outcomes like positive sentiment, actual recommendations, traffic generation, or conversion rates.

Brand Radar offers two primary methods for measuring AI visibility: exploring a broad index of pre-collected AI responses or tracking a custom set of user-defined prompts. The broad index provides immediate insights by analyzing existing data, while tracked prompts begin data collection upon their addition. The accuracy and meaningfulness of these results depend on clearly defined entities, filters, and locations.

To begin, users add their brand and competitors, specifying names and associated websites. This allows for grouping variations, rebrands, or product names under a single entity. The "Explore" function utilizes a pre-existing corpus of AI responses to generate immediate reports on mentions, citations, estimated impressions, and AI SOV. This view calculates AI SOV as the brand’s share of estimated impressions within the filtered corpus, relative to tracked competitors.

Users can then refine their analysis using report filters to define the market and platform scope. Selecting specific AI platforms like ChatGPT, Gemini, Perplexity, Copilot, Google AI Overviews, and Google AI Mode is critical, as brand performance can vary significantly across them. Examining platform-specific data can reveal weaknesses that might be masked in combined scores. Saved reports allow users to revisit specific configurations, and these can be mapped to Ahrefs projects for ongoing tracking.

For more targeted monitoring, "Tracked prompts" enable users to focus on commercially important questions. Users can add custom prompts, specify locations, and set AI assistants and update frequencies. Prompt sets can be generated or manually created, focusing on relevant groups of questions that address customer concerns.

The Overview report presents key metrics side-by-side: mentions, citations, estimated impressions, and AI SOV, each answering different questions. Mentions indicate direct brand presence, citations show which websites AI references, and estimated impressions quantify potential exposure. Users can drill down into AI responses, cited domains, and cited pages for deeper analysis. It is vital to maintain measurement consistency. Altering prompts, competitors, or platforms invalidates like-for-like comparisons. Changes to the measurement setup should be meticulously documented.

The interpretation of AI SOV requires careful consideration of context. A 20% SOV has different implications depending on the number of competitors and the specific prompt set used. Comparing a brand’s AI SOV against its competitors provides a more meaningful benchmark than generic industry averages. Similarly, tracking a brand’s SOV against itself over time, while maintaining consistent measurement parameters, reveals genuine trends. Industry benchmarks should be used cautiously, as they may not reflect the specific scope and methodology of a particular measurement.

An apparent change in AI SOV can stem from three primary factors: genuine changes in visibility, modifications to the measurement setup, or alterations within the AI systems themselves. If the prompts, competitors, and platforms remain constant, an increase in SOV suggests an improvement in brand visibility. Conversely, changes in the measurement parameters mean the ruler has been altered, not necessarily the visibility. Furthermore, AI models are dynamic; updates, retrieval changes, or inherent variability can lead to fluctuations in responses that do not reflect a decline in a brand’s actual SEO or content performance. Trends that persist across multiple reporting periods are more reliable indicators than isolated spikes.

As AI search technology matures, AI SOV is expected to become a more refined and useful metric. Improvements in data collection, sampling, demand estimation, and the ability to distinguish meaningful visibility from passing mentions will enhance its value. However, the fundamental challenge of defining the measurement scope will remain. The true advantage will lie in establishing a trusted measurement framework, monitoring changes over time, and understanding the underlying reasons for those shifts. AI SOV, when used as a tool for understanding evolving brand positioning in AI responses rather than just another score to chase, can become genuinely valuable.