Tracking • informational intent
Share of Model Voice: How to Measure Your Brand's Presence Across AI Platforms
Share of Model Voice (SOMV) is the metric that tracks how often AI models recommend your brand versus competitors. Learn how to measure it, what it reveals, and how to improve your position across ChatGPT, Claude, Gemini, Grok, and Perplexity.
The metric traditional analytics misses
You can track organic search traffic, paid ad clicks, and social media mentions. But when a user asks ChatGPT 'what's the best CRM for freelancers?' and the model recommends your competitor instead of you — that interaction never shows up in your analytics. There's no click, no impression, no referral. Just a recommendation that shaped a buying decision you'll never know about.
Share of Model Voice (SOMV) is the metric designed to capture this invisible influence. It measures how often AI models mention, recommend, or cite your brand compared to competitors for specific queries in your market. Think of it as Share of Voice for the age of AI-generated answers.
As AI assistants become a primary research channel (an estimated 40% of product research queries now go through AI chatbots before or instead of Google), the brands that measure and optimize SOMV will have a structural advantage over those still focused exclusively on traditional search.
How SOMV works
SOMV measurement works by sampling AI models with the queries your target customers actually ask. For a project management tool, that might be queries like 'best project management software for remote teams', 'Asana vs Monday.com alternatives', or 'free task management tools with Gantt charts'.
For each query, the system records which brands the AI model mentions, in what context (recommendation, comparison, passing mention), and how prominently (first recommendation vs. last in a list). This data is collected across multiple AI platforms — ChatGPT, Claude, Gemini, Grok, and Perplexity — because each model has different training data and retrieval sources, resulting in different recommendation patterns.
The result is a visibility score per platform per keyword, plus an aggregate view of your brand's overall AI presence compared to your competitive set.
What SOMV reveals that other metrics don't
SOMV answers questions that no other metric can.
- Model-specific blind spots — You might be well-known to ChatGPT but invisible to Perplexity, or strong in Claude's training data but missing from Gemini. Each platform has different data sources and biases. SOMV reveals exactly where you're strong and where you're absent.
- Keyword-level positioning — For some queries, you might be the top recommendation; for others, you don't appear at all. SOMV maps your visibility across your entire keyword portfolio, revealing which queries you dominate and which ones competitors own.
- Competitive movement — When a competitor publishes better structured data, improves their content, or gets cited by more authoritative sources, their SOMV rises — often at your expense. Tracking this over time shows competitive shifts before they manifest in traditional traffic metrics.
- Content effectiveness — After you improve your structured data, rewrite key pages, or publish llms.txt, SOMV shows whether those changes actually made your site more likely to be recommended. This closes the feedback loop between AI visibility work and measurable outcomes.
How to set up SOMV tracking
Effective SOMV tracking requires three components: the right keywords, the right competitive set, and a consistent sampling schedule.
Start with 5-10 keywords that represent your core buying queries — the questions a potential customer asks when evaluating solutions in your category. Avoid vanity keywords (your brand name) and focus on category and comparison queries where AI recommendation matters most.
Define your competitive set: the 3-5 brands that appear most frequently alongside you in AI answers. This set may differ from your traditional SEO competitors because AI models have different knowledge distributions.
Sample weekly at minimum. AI model behavior changes as models are retrained, retrieval indexes are updated, and web content evolves. A single snapshot is interesting; a trendline over weeks is actionable.
Interpreting your SOMV data
A visibility score of 70+ on a platform means you're consistently mentioned and well-positioned for that keyword. Below 30 means you're rarely or never mentioned. The interesting range is 30-70, where small improvements in content structure, structured data, or authority can meaningfully shift your position.
Pay attention to the gap between heuristic SOMV (estimated from your site's technical readiness) and live SOMV (what the model actually says). A high heuristic score but low live score suggests your site is technically ready but the model hasn't incorporated your content yet — often due to recent changes that haven't propagated through retraining or retrieval index updates.
A low heuristic score with a high live score is rarer but indicates brand strength — the model recommends you based on training data authority despite technical issues on your site. This is a fragile position because retrieval-based systems will eventually deprioritize technically inferior content.
Improving your SOMV
SOMV improvements come from two directions: making your site more accessible to AI models (technical) and making your content more citable (authority).
On the technical side, the same AI visibility fundamentals apply: allow AI crawlers, publish llms.txt, add structured data, ensure content is server-rendered. These changes improve your heuristic score and make your content available for retrieval-based answers.
On the authority side, the game is longer-term. Get mentioned in industry publications, earn backlinks from authoritative domains, publish original research and data, and create the definitive resource for your category. AI training data reflects the web's link and citation graph — the more authoritative sources reference you, the more likely models are to recommend you.
The most actionable short-term lever: create specific, factual comparison and FAQ content that directly answers the queries you're tracking. AI models strongly prefer pages that explicitly address the question being asked over pages that are tangentially relevant.
Execution Checklist
- • Identify 5-10 core buying queries where AI recommendation matters for your business.
- • Define your competitive set (3-5 brands that appear in AI answers alongside you).
- • Run initial SOMV baseline across all five major AI platforms.
- • Compare heuristic score (technical readiness) vs. live score (actual model output) to identify gaps.
- • Prioritize fixes on platforms where you have the largest gap between potential and actual visibility.
- • Set up weekly sampling to track trends and competitive movement.
- • After each round of AI visibility improvements, measure SOMV change to validate impact.
FAQ
How is SOMV different from traditional Share of Voice?
Traditional Share of Voice measures your brand's presence in search results, ads, or media mentions — all channels where impressions and clicks are trackable. SOMV measures your presence in AI-generated answers where there are no impressions or clicks to count. The only way to measure SOMV is to directly query AI models and analyze their responses.
Which AI platforms should I track SOMV on?
Start with the five major platforms: ChatGPT (OpenAI), Claude (Anthropic), Gemini (Google), Grok (xAI), and Perplexity. Each has different training data and retrieval sources, so your visibility varies across them. If your audience skews toward a specific platform (e.g., developers tend to use Claude and ChatGPT more), weight those platforms higher in your analysis.
Can a small brand compete with established players in SOMV?
Yes, especially on retrieval-based platforms like Perplexity and ChatGPT Browse. These systems fetch pages in real time, so a small brand with excellent structured data, a clear llms.txt, and specific factual content can appear alongside much larger competitors. Training-based SOMV is harder to influence quickly because it depends on the broader web's citation graph, but retrieval-based visibility can shift within days of making technical improvements.