How Do I Measure Share of Model?

If you've been tracking your brand's Google rankings for years and suddenly feel like the ground is shifting under your feet, you're not imagining it. ChatGPT now handles more than 1 billion queries per day, and a growing chunk of your potential customers are getting answers directly from AI - without ever clicking a link. That changes everything about how you measure brand visibility. Share of Model is the metric that tries to make sense of this new reality, and once you understand how to measure it, it starts to feel less scary and more like an opportunity.

What Share of Model Actually Means

Think of it as your brand's presence score inside the minds of AI systems. When someone asks an AI "what's the best project management software for remote teams?" or "which accounting firm should I use in London?", does your brand show up? And how often, compared to your competitors? That frequency, expressed as a percentage, is your Share of Model.

It's fundamentally different from anything we've tracked before. It's not about page rankings or keyword positions. It's about whether an AI system has absorbed enough about your brand to confidently recommend or mention you when responding to relevant questions.

Understanding what Share of Model is in AI search is essential before diving into measurement. The concept is genuinely different because it measures something SEO never could: how familiar AI systems are with your brand and whether they consider you credible enough to cite.

Why Traditional Metrics Are No Longer Enough

Here's the uncomfortable truth: the metrics most marketing teams still rely on were built for a world that's quietly disappearing. Over 60% of informational B2B queries now end without a click to a website. That's not a rounding error - that's the majority of your potential audience getting their answer from an AI summary and moving on.

Traditional Share of Voice measures how loudly you're talking relative to competitors across paid and organic channels. It's useful, but it tells you nothing about what an AI system thinks about your brand when someone asks it a direct question.

Google's AI Overviews alone dropped click-through rates for top-ranking pages by 34.5% in one study. So even if you rank number one organically, you might be getting far less traffic than you used to. The visibility that matters is increasingly happening inside AI-generated answers, not below them.

This is why tracking AI visibility as a core KPI has become critical. It's not a future concern. It's a present one.

The Step-by-Step Process for Measuring SoM

Measuring Share of Model requires discipline, but it's straightforward once you understand the process:

1. Define your relevant prompts. Write out the questions your target customers would realistically ask an AI. These should cover your product category, use cases, and the problems you solve. Aim for at least 100 prompts to get a statistically meaningful sample across different query intent types.
2. Test across multiple AI platforms. Run each prompt through ChatGPT, Gemini, Perplexity, Claude, and any other platform your audience uses. Each model has different training data and retrieval logic, so your SoM will vary significantly across platforms. Understanding where each platform sources its information helps explain these differences.
3. Record every brand mention. For each response, note which brands are mentioned, recommended, or cited. Include your brand and your main competitors. Track whether mentions are recommendations, passing references, or comparisons.
4. Calculate your percentage frequency. Divide the number of responses that include your brand by the total number of prompts tested, then multiply by 100. Do the same for each competitor. That gives you a comparative SoM score.
5. Track over time. A single snapshot is interesting. A trend line is actionable. Run the same prompt set monthly to see whether your SoM is growing, shrinking, or staying flat relative to competitors.

The average SoM for category leaders is approximately 34%, with improvements of 10% or more typically correlating with a 28% increase in AI-attributed inquiries, according to Steakhouse - Share of Model Framework.

Platform-Specific Performance Differences

Your SoM is not a universal score - it varies dramatically across platforms. This variation is itself valuable data:

Perplexity tends to produce the highest concentration of brand citations because its retrieval-augmented generation architecture explicitly surfaces and attributes sources, making citations more frequent and transparent.

ChatGPT generates fewer explicit citations due to its training approach, but when your brand does appear, it often carries significant weight given the platform's 815 million monthly active users.

Google Gemini has its own citation patterns influenced by how Google indexes and ranks content, often favoring large established brands but showing less diversity than other platforms.

A brand might command 24% of mentions on one model but less than 1% on another. This tells you where your content is being picked up and, conversely, where you're missing opportunities.

What Affects Your Share of Model Score

AI systems don't just pull brand names out of thin air. They surface brands that appear consistently, authoritatively, and clearly across the sources they've been trained on or can retrieve from.

"Share of Model measures a brand's presence within AI data sets, specifically LLMs, as a proportion of the total mentions within a category. It's the metric that tells you whether the AI has heard of you, and whether it thinks you're relevant to the questions its users are asking."

Several factors directly influence your SoM:

• Consistency of brand descriptions. If your website says one thing about what you do, your LinkedIn says something slightly different, and your press releases use different terminology again, AI systems get confused. Inconsistent brand descriptions across platforms reduce AI confidence and negatively impact your SoM. Standardize your brand positioning.
• Volume and quality of mentions across authoritative sources. The more high-authority sources that mention your brand in relevant contexts, the more likely AI systems are to include you. This includes news coverage, industry publications, research citations, and credible review sites. LLMs have a "citation bias" - they prefer sources with data, statistics, and clear definitions.
• Original, extractable content on your own site. AI systems reward content that directly answers questions with original information. Vague, generic, or jargon-heavy pages don't get cited. Content with unique data points, proprietary research, and clear entity relationships gets cited more often. Structured data and proper heading hierarchy also improve extractability.
• Citations and third-party validation. Being referenced by credible external sources matters significantly. Semrush - AI SEO Tips notes that earning citations in authoritative sources is one of the most effective ways to improve AI visibility.
• Entity relationships and semantic clarity. AI models think in entities and relationships. If your brand (Entity A) is clearly associated with "enterprise security" (Entity B) in high-authority contexts, the model learns this relationship and surfaces you when that topic comes up.

Building Your SoM Improvement Strategy

Create original, citable content. Publish original research, white papers, and definitive guides rather than generic how-to content. Content with proprietary data, first-hand examples, and unique perspectives is more likely to be cited by AI systems that prioritize information beyond what's already widely available.

Optimize for AI extractability. Use structured data (Schema.org markup), clear heading hierarchies, concise definitions, and information-rich paragraphs. Define who you are, what you do, and who you serve using language machines can easily parse.

Build third-party citations. Start with earned media - pitch stories to industry publications, contribute to analyst reports, and get quoted by credible sources. These citations signal authority to AI systems and improve your SoM across multiple platforms.

Ensure brand consistency. Standardize how your brand is described across your website, LinkedIn, press releases, and other touchpoints. Consistency helps AI systems build confidence about what you do.

SoM as Part of Your Broader KPI Framework

Share of Model doesn't replace your existing metrics - it adds a critical layer that the others can't cover. Think of it as one piece of a broader KPI framework for AI search.

Here's how SoM fits alongside other metrics:

• Share of Voice tells you how much of the conversation in your category you own across paid and earned media relative to competitors. Learn how to measure Share of Voice to complement your SoM tracking.
• AI-driven traffic tells you how many people are actually clicking through from AI platforms to your site - the downstream conversion metric.
• Brand mentions in AI responses tells you how often you appear, even when no click happens - pure visibility.
• Share of Model ties it all together by giving you a competitive percentage that shows your relative standing within each AI system.

Together, these give you a complete picture of your brand's visibility in an AI-first environment.

Tracking Share of Model: Tools and Approaches

Manual testing works for small samples, but for consistent tracking, most brands eventually move to dedicated tools. Lumentir tracks ChatGPT, Grok, Gemini, Google AI Overviews, Copilot, and Perplexity separately, enabling SoM measurement per platform automatically. The entry plan covers 1 website, 3 topics, 100 prompts, and up to 3,000 responses per month - enough to establish baseline SoM across multiple platforms and track monthly trends.

The key advantage of automated tracking is consistency. When you test the same 100 prompts monthly across 6 platforms, you get reliable trend data that reveals whether your SoM is growing, stagnating, or declining - and on which platforms you're gaining or losing ground.

Common Mistakes When Measuring SoM

Teams often undermine their own results with a few avoidable errors:

• Testing too few prompts. Ten or twenty prompts is not enough. AI responses vary significantly based on phrasing and context. Aim for at least 100 prompts across different intent types and question formats.
• Only testing one AI platform. Your SoM on ChatGPT might be 40% while on Perplexity it's 15%. That gap tells you something important about where your content is and isn't being picked up. Multi-platform testing is essential.
• Ignoring competitor data. SoM is a relative metric. Knowing you appear in 30% of responses means little unless you know your main competitor appears in 60%. Always track competitors.
• Measuring once and moving on. AI models update, new content gets indexed, and your competitors are actively working on their own visibility. Monthly tracking is the minimum cadence.
• Not connecting SoM to strategy. If your score is low, the fix isn't mysterious. It usually means your content isn't answering the right questions clearly enough, or you need more third-party citations and mentions.

What a Good SoM Score Looks Like

Honestly, there's no universal benchmark yet because the field is still new. But here's a practical framework: if you're a market leader in your category and you're appearing in fewer than 30% of relevant AI responses, that's a problem worth addressing. If you're a smaller brand and you're appearing in 20% of responses while the category leader appears in 50%, that gap is your roadmap.

The goal isn't to appear in every response. It's to appear consistently in responses where your brand is genuinely relevant, and to appear more often than your direct competitors for the queries that matter most to your business.

"AI-generated responses influence brand perception, authority ranking, competitive positioning, user trust, and downstream search behavior. That's a lot riding on a metric most marketing teams aren't tracking yet. Which, if you think about it, is actually good news for the ones who start now."