Share of Model: The AI Marketing Metric Replacing Share of Voice

Share of model is the percentage of relevant AI-generated responses that mention, recommend, or cite your brand. It’s the AI-era equivalent of share of voice — but instead of measuring how much of the advertising conversation you own, it measures how much of the AI recommendation layer you own.

The calculation is straightforward. You define a set of buying-intent prompts that represent how your customers discover products in your category. You send each prompt to multiple AI models (ChatGPT, Perplexity, Gemini, Claude). You count how many responses mention your brand. Share of model = (responses mentioning your brand / total responses tested) × 100. If you test 15 prompts across 4 models (60 total tests) and your brand appears in 9 responses, your share of model is 15%.

But raw mention rate is only part of the picture. There are three tiers of visibility within share of model. Mentioned means your brand name appears somewhere in the response. Recommended means the AI actively suggests your brand as an option worth considering — this is more valuable than a passing mention. Cited means the AI includes a link to your website or a specific page. Track all three, but weight your analysis toward recommendation rate. A brand that’s mentioned in 20% of responses but recommended in only 2% has a visibility problem that raw mention rate masks.

According to Joel House, founder of MentionLayer and author of AI for Revenue, "Share of model is to the AI era what share of voice was to the advertising era — except it’s more measurable, more actionable, and more directly correlated with revenue. Every percentage point of share of model you gain represents real customers being directed to you instead of your competitors."

Why does this matter more than traditional metrics? Because AI recommendations have disproportionate conversion power. When a trusted AI model recommends your brand, it carries the weight of an expert endorsement. In our experience, AI-referred visitors convert at a meaningfully higher rate than organic search visitors. They arrive with higher intent and higher trust because an AI model essentially pre-qualified and pre-sold them. Share of model directly predicts this AI-referred revenue. Every percentage point of share of model you gain translates to a measurable increase in high-intent traffic and conversions.

Share of voice measured how loud your brand was in a noisy room. Share of model measures whether the most trusted advisor in the room is recommending you. If you haven’t tested your AI visibility yet, start with the 60-second AI visibility test. In a world where 37% of consumers start with AI, that trusted advisor is ChatGPT, Perplexity, or Gemini. If they’re not recommending you, the most qualified, highest-intent prospects in your market are being directed to your competitors before they ever know you exist.

Here’s a step-by-step process you can implement today, with no special tools required.

Step 1: Choose 10-15 buying-intent prompts. These should represent how real customers ask about your category. Don’t use your brand name. Use the language your customers use. For a project management tool: "What’s the best project management software for small teams?" "Can you recommend a Trello alternative?" "Compare the top project management tools in 2026." "I need a kanban board tool for my startup — what’s good?" "What project management app has the best free tier?" Mix specific comparison queries with broad recommendation queries. Include queries that mention competitors ("alternatives to X") because these are high-intent moments where your brand needs to appear.

Step 2: Test across 4 AI models. Open ChatGPT, Perplexity, Gemini, and Claude. Run each prompt in each model. For ChatGPT, use GPT-4 with browsing enabled for the most representative results. For Perplexity, use the default mode. Each platform has different citation behaviours, which we break down in our platform-by-platform GEO guide. For each response, record: (a) Does your brand appear? (b) Is it recommended as an option? (c) Is your URL cited? (d) What position is it in — first mention, middle, last? (e) Which competitors appear? (f) What sources are cited?

Step 3: Calculate your scores. Create a simple spreadsheet with prompts as rows and models as columns. Mark each cell as: 0 (not mentioned), 1 (mentioned), 2 (recommended), 3 (recommended + linked). Your mention rate is the percentage of cells scoring 1 or higher. Your recommendation rate is the percentage scoring 2 or higher. Your citation rate is the percentage scoring 3. These three numbers are your share of model profile.

Step 4: Track weekly. Use the exact same prompts every week for consistency. Small variations in AI responses are normal — what you’re tracking is the trend, not any single test. Over 4-6 weeks, clear patterns emerge. Some prompts consistently include you; others consistently don’t. The prompts where you’re invisible are your optimization targets. Log the date, the scores, and any changes in your GEO activities so you can correlate actions with results.

Here are example prompts you can adapt for different industries: - SaaS: "What are the best [category] tools for [audience] in 2026?" - Professional Services: "Who are the best [service type] firms in [city/region]?" - E-commerce: "What’s the best [product type] for [use case]?" - B2B: "Can you compare the top [category] vendors for [company size]?" - Local: "Who do people recommend for [service] near [location]?"

After measuring share of model across hundreds of brands, clear benchmark tiers have emerged. These aren’t arbitrary — they correspond to meaningfully different levels of AI-referred business impact.

0-5%: Invisible. Your brand appears in fewer than 1 in 20 AI responses for your category. This is where 65.9% of businesses sit (n=1,004), according to our AI Visibility Index study across 1,004 businesses and 95,392 data points. At this level, AI-driven discovery is sending essentially zero traffic to your business. You’re losing every AI-influenced buying decision by default. The good news: moving from 0% to 5% is achievable within 30-60 days with focused citation seeding.

5-15%: Emerging. Your brand occasionally appears in AI recommendations. You’re showing up on some platforms for some queries, but inconsistently. At this level, you’re starting to capture some AI-referred traffic, but you’re not a reliable recommendation. Think of it as the AI equivalent of ranking on page two of Google — you exist, but you’re not where the action is. Most brands can reach this tier within 60-90 days of starting a GEO program.

15-30%: Competitive. Your brand is a regular part of the AI recommendation set for your category. You appear across multiple models and multiple prompt variations. At this level, you’re capturing meaningful AI-referred traffic and conversions. You’re in the consideration set alongside the category leaders. This is where most well-executed GEO programs arrive within 6 months.

30%+: Category Leader. Your brand appears in more than 1 in 3 AI responses for your category. You’re the first or second brand mentioned in most responses. At this level, you’re capturing a disproportionate share of AI-referred revenue. This position is self-reinforcing: AI models develop strong associations between your brand and your category, making it increasingly difficult for competitors to displace you. Reaching this tier typically requires 6-12 months of comprehensive GEO work across all 6 pillars.

"In our experience running AI visibility campaigns at MentionLayer, we’ve found that the jump from 0% to 15% share of model is the hardest — it requires building your brand’s presence from scratch in the AI consensus layer," says Joel House. "But once you cross 15%, each additional percentage point comes easier because AI models start reinforcing your position organically."

Two critical notes on benchmarks. First, the exact number matters less than the trend. AI responses have natural variability. Your share of model might be 12% one week and 15% the next without any change in your GEO activities. What matters is the 4-week rolling average and the direction it’s moving. Second, benchmarks vary by category competitiveness. In a niche B2B category with 5 competitors, 30% share of model means you’re dominating. In a consumer category with 50 competitors, 10% share of model might make you the leader. Always contextualize your score against the competitive landscape.

There are five primary levers for improving share of model, each corresponding to a pillar of the AI visibility framework. The right lever to pull first depends on your current score and competitive position.

Lever 1: [Citation Seeding](/blog/what-is-llm-seeding) (fastest impact). Place authentic, helpful responses in high-authority Reddit threads, Quora answers, and forum discussions that AI models already reference. This is the highest-ROI GEO tactic because it directly inserts your brand into the information sources AI models cite. Target threads that rank on Google page one for your keywords, have high engagement (20+ comments), and are already cited by AI models. A focused campaign of 3-5 quality placements per week typically moves share of model by 5-10 percentage points within 60-90 days. See the Citation Seeding Playbook for the full methodology.

Lever 2: [Entity Optimization](/blog/entity-seo-knowledge-graph) (foundational). Clean up your brand’s entity data across all platforms. Ensure consistent descriptions, categories, and contact information on Google Business Profile, LinkedIn, Crunchbase, and relevant directories. Implement comprehensive schema markup on your website: Organization, Product, FAQ, Review, and BreadcrumbList. Entity optimization doesn’t directly increase mentions, but it increases the probability that mentions translate into recommendations. AI models are more likely to recommend brands they can confidently classify and describe.

Lever 3: Review Volume (trust signal). AI models use review data as a trust signal when deciding whether to recommend a brand. A brand with 500 reviews and a 4.3 average is more likely to be recommended than a brand with 12 reviews and a 4.8 average — because the volume signals validated market presence. Focus on the platforms AI models check most: Google Reviews, Trustpilot, G2, and Capterra. A review velocity of 10+ new reviews per month across platforms is the target.

Lever 4: [Press Coverage](/blog/digital-pr-ai-era) (authority signal). Earned media in authoritative publications signals to AI models that your brand is established and credible. This doesn’t require coverage in the New York Times — industry-specific publications often have more weight for category-specific queries. Target 2-3 press placements per month in publications relevant to your space. Focus on publications that AI models actually cite (check Perplexity’s sources for your category to identify them).

Lever 5: Content Optimization (supporting signal). Ensure your own website content is optimized for AI consumption. This means first-person experience-based writing (1.67x citation improvement), comprehensive FAQ pages, detailed product comparison content, and fresh content updated within the last 30 days. Your website content is the 10% that brand-owned sources contribute to AI understanding — make it count.

"Most brands try to boil the ocean with GEO. The data says start with Reddit citation seeding and entity cleanup — those two levers alone can move you from invisible to competitive within 90 days," says Joel House.

Here’s how to prioritize based on your current share of model: - 0% share of model: Start with Lever 1 (citations) + Lever 2 (entities). These move the needle fastest. - 5-15% share of model: Continue Lever 1, add Lever 3 (reviews) and Lever 4 (press) to build authority. - 15-30% share of model: All five levers running in parallel. Focus on Lever 4 (press) to break into the top tier. - 30%+ share of model: Defend your position with continued Lever 1 activity and aggressive Lever 4 to maintain authority advantage.

Your share of model in isolation tells you half the story. The other half is how you compare to competitors. Track competitor share of model using the exact same prompts and models — the only difference is you’re counting their brand mentions instead of yours.

Build a share of model leaderboard for your category. For each competitor, record their mention rate, recommendation rate, and citation rate across your prompt set. Rank them. Identify who’s winning and on which platforms. You’ll typically find that different competitors dominate different models. One competitor might own ChatGPT recommendations but be weak on Perplexity. Another might dominate Perplexity through heavy Reddit presence but barely register on Gemini. These gaps are strategic opportunities.

The gap analysis is where competitive intelligence becomes actionable. For each prompt where a competitor appears but you don’t, ask: What information is the AI model drawing on? Check the cited sources. Is the competitor mentioned in Reddit threads you’re absent from? Do they have press coverage you don’t? Is their review volume dramatically higher? Each gap maps to a specific pillar and a specific action. If a competitor appears in Perplexity responses citing three Reddit threads where they’re mentioned and you’re not — those are your citation seeding targets.

Track competitor trends alongside your own. If a competitor’s share of model is growing at 3% per month while yours grows at 1%, you’re falling behind even though your absolute number is improving. Conversely, if you’re growing at 5% while the category leader is flat, you’re on a trajectory to overtake them. The relative rate of change matters as much as the absolute score.

One powerful use of competitive share of model data is identifying market positioning opportunities. If every competitor is described by AI models as a "project management tool for teams," but none are described as a "project management tool for freelancers," you’ve found a positioning gap. Seed content that positions your brand specifically for the underserved angle. AI models are responsive to specific, niche positioning because it makes their recommendations more useful and differentiated. The brand that owns a specific niche within a category often has higher share of model for niche-specific prompts than the category leader has for broad prompts. For tools that can automate this tracking, see our AI visibility tools comparison. MentionLayer’s citation engine automates competitive share of model tracking alongside discovery and seeding, giving agencies a complete workflow from insight to action.

Want your starting share of model before you build the leaderboard? A free AI visibility audit measures your brand against your category’s competitors across every major AI model and emails the numbers back in about 20 minutes.