Share of Model: The AI Visibility Metric 45% of Marketers Can't Even Measure Yet

AI platforms now field 45 billion sessions per month globally — a volume equal to 56% of all traditional search traffic, according to Similarweb. Every one of those sessions represents a moment where an AI might recommend your brand, recommend a competitor, or skip your category entirely. Share of Model is the metric that quantifies which of those three outcomes is actually happening.

The problem is that most marketing teams have no idea where they stand. Semrush data shows 45% of marketing leaders cannot accurately measure their brand's AI visibility, and only 9% possess tools capable of doing so. Meanwhile, 43% of marketers name AI visibility as a core strategy priority — but only 14% actually track AI citations, according to Digital Applied research.

This guide breaks down what Share of Model measures, how the math works, why AI recommendations are far less stable than most teams assume, and what marketing managers, analytics teams, and brand strategists can do about the measurement gap right now.

What Is Share of Model and Where Did the Metric Come From?

Share of Model (SOM) measures the proportion of times an AI platform mentions a brand out of all brand mentions in the same product category — giving marketing teams a concrete number for brand visibility inside AI-generated answers. Jack Smyth, Chief Solutions Officer for AI at Jellyfish, and Tom Roach formalized the concept at Jellyfish (part of The Brandtech Group) in late 2024.

Symphonic Digital defines Share of Model as "the number of mentions of a brand by one or multiple LLMs, as a proportion of total mentions of brands in the same category." Jellyfish subsequently launched a commercial "Share of Model" trademarked platform, with Danone and Pernod-Ricard's Chivas Brothers unit as beta clients.

The core formula is straightforward:

(Your brand mentions / Total category mentions) x 100

If an AI platform mentions your CRM tool 30 times across 100 category-related prompts while mentioning all CRM tools a combined 200 times, your Share of Model is 15%.

The metric goes by several names across the industry. Analysts and vendors variously call it "AI Share of Voice," "Share of Citation," "Share of Answer," and "SoLLM." The industry appears to be consolidating around "AI Share of Voice" as the standard term, though the underlying calculation remains the same regardless of branding.

Jack Smyth framed the stakes in direct terms: "LLMs are no longer just tools; they are a critical part of the customer journey and an audience." Catherine Lautier, Vice President at Danone, described the platform's value as enabling teams to compare each LLM's perception of brand ecosystems — understanding not just whether a brand appears, but how different AI models contextualize the brand relative to competitors.

For marketing managers allocating budget across channels, Share of Model answers a question that traditional analytics tools miss entirely: when a prospect asks ChatGPT or Perplexity "what's the best email marketing platform for ecommerce," does your tool appear in the answer? Tools like Google Analytics can track referral traffic from AI platforms, but SOM measures the upstream visibility that drives those referrals in the first place.

For analytics teams building measurement frameworks, Share of Model fills the attribution gap between brand awareness campaigns and AI-referred conversions. Similarweb data shows AI-referred visitors convert at 7.1% — second only to paid search at 7.8% — which means the traffic SOM influences carries genuine commercial value.

For brand strategists managing competitive positioning, Share of Model reveals whether AI platforms perceive your brand the way your positioning strategy intends. A brand investing heavily in the "enterprise" segment might discover through SOM tracking that AI models overwhelmingly recommend the brand for "small business" use cases instead.

How Does Share of Model Differ from Share of Voice and Share of Search?

Share of Model measures brand presence within AI-synthesized answers, which are fundamentally different from the multi-result feeds of traditional search or the rented impressions of advertising. Share of Voice tracks paid advertising presence, Share of Search tracks branded search query volume, and Share of Model tracks mentions inside AI-generated responses.

The distinctions matter because each metric reflects a different kind of brand presence:

• Share of Voice measures advertising presence across channels — the proportion of total ad impressions or spend a brand captures in its category. Share of Voice is "rented" through media budgets and disappears the moment ad spend stops. A brand with a $500,000 monthly ad budget buying 25% of available impressions has a 25% Share of Voice
• Share of Search measures the proportion of branded search queries a brand captures relative to competitors. Popularized by Les Binet at the IPA, Share of Search reflects organic consumer interest — when people type "Semrush" versus "Ahrefs" versus "Moz" into Google, the relative volume indicates brand consideration. Our Ahrefs review covers how the platform tracks branded search metrics
• Share of Model measures brand presence inside AI-generated synthesized answers. Unlike Share of Voice, SOM cannot be bought through ad spend — AI platforms decide which brands to mention based on training data, retrieval sources, and recommendation logic. Unlike Share of Search, SOM captures what AI platforms recommend, not what consumers search for

The practical difference is clearest in the implications for marketing teams. Improving Share of Voice requires budget. Improving Share of Search requires sustained brand-building campaigns. Improving Share of Model requires a different playbook entirely — one focused on structured data, citation-worthy content, and presence across the sources AI platforms retrieve from.

Semrush analyzed 126 million prompts across major AI platforms from January through April 2026 and found that only 36 global brands maintained top-100 visibility across all four major AI platforms every month — a group researchers labeled the "Universal 36." That extreme concentration suggests Share of Model follows winner-take-most dynamics even more aggressive than traditional search ranking. Teams using Semrush for traditional SEO monitoring are increasingly pairing that data with AI visibility tracking to understand both halves of the discovery equation.

Why Are AI Brand Recommendations So Wildly Inconsistent?

AI brand recommendations show extreme variability — SparkToro's January 2026 study of 600 volunteers and 2,961 prompts found less than a 1-in-100 chance of getting the same brand list twice from ChatGPT, and roughly a 1-in-1,000 chance of getting the same list in the same order. Ranking position within any single AI response is effectively random.

SparkToro founder Rand Fishkin put the finding bluntly: "AIs do not give consistent lists of brand or product recommendations." Fishkin explicitly warned that any tool claiming to provide reliable ranking position data is misleading — the meaningful metric is frequency of occurrence across 60 to 100 or more independent runs, not placement in any single response.

Cross-platform fragmentation compounds the inconsistency. Research shows only 11% domain overlap between ChatGPT and Perplexity citations, and only 2.1% of Google's top-10 organic pages appear among ChatGPT's citations. A brand that ranks well in traditional search results has no guarantee of appearing in AI-generated answers.

The fragmentation is not just between platforms but within them. Consider the documented examples from cleaning product brands: Ariel detergent held 24% Share of Model on Meta Llama but less than 1% on Google Gemini. Chanteclair captured 19% SOM on Perplexity but 0% on Llama. The same product category, the same market — radically different brand visibility depending on which AI platform the consumer happens to use.

The broadest finding is perhaps the most sobering: 52% of brands never surface at all across AI engines. More than half of all tracked brands do not appear in AI-generated recommendations in any meaningful volume on any platform.

For analytics teams running measurement programs, the SparkToro data means any Share of Model tracking methodology must aggregate across large sample sizes. A single prompt to ChatGPT asking "what's the best analytics tool" tells a team almost nothing — the answer might name Amplitude one time and Mixpanel the next. Tracking tools need to run hundreds of prompts and measure mention frequency, not snapshot rankings. Our Amplitude vs Mixpanel comparison covers how both platforms approach data analysis, but neither platform's own analytics can yet track how often AI recommends them.

For brand strategists managing multi-market portfolios, the Ariel and Chanteclair data reveals that SOM strategy must be platform-specific. A unified "AI optimization" initiative is insufficient when each AI model draws from different training data, different retrieval sources, and different recommendation algorithms.

The industry currently spends an estimated $100 million or more annually on AI tracking tools — a figure that reflects how seriously brands take the measurement challenge, even as the underlying metrics remain volatile.

How Do Different AI Platforms Cite Sources Differently?

Each major AI platform follows distinct citation patterns — ChatGPT leans on third-party directories, Gemini favors brand-owned websites, Perplexity emphasizes reviews, and Google AI Overviews direct traffic most heavily to brand domains. Profound's analysis of 6.8 million citations provides the clearest picture of these structural differences.

The platform-by-platform breakdown reveals why a single content strategy cannot optimize Share of Model universally:

• ChatGPT includes approximately 6 to 8 citations per response, with 48.73% of cited sources coming from third-party directories. Brands that maintain accurate listings on aggregator sites, review platforms, and industry directories gain disproportionate visibility in ChatGPT responses
• Google Gemini shows wider citation variability at 3 to 17 citations per response, with 52.15% coming from brand-owned websites. Gemini's reliance on first-party sources means brands with comprehensive, well-structured websites have a structural advantage on this platform
• Perplexity generates the most citation-dense responses at approximately 21.87 citations per answer, with an emphasis on review content. Brands with strong profiles across review platforms (G2, Capterra, Trustpilot) benefit from Perplexity's retrieval patterns
• Google AI Overviews direct 59.8% of citations to brand domains — the highest first-party rate of any major platform. AI Overviews function more like enhanced search results than synthesized recommendations, which explains the brand-domain bias

One cross-platform constant stands out: Reddit is cited at approximately 40% frequency across all AI engines. Reddit's combination of authentic user discussion, specific product comparisons, and detailed experience reports makes its content unusually attractive to AI retrieval systems regardless of platform.

For marketing managers planning content distribution, the Profound data provides a clear allocation framework. Investing in third-party directory presence pays off most on ChatGPT. Investing in owned-website content pays off most on Gemini and AI Overviews. Investing in review generation pays off most on Perplexity. And investing in Reddit presence — through genuine community participation, not astroturfing — provides a baseline lift across every platform.

This platform-specific citation pattern also explains why tracking AI visibility requires multi-platform measurement. A brand might hold strong SOM on Gemini (where its website ranks well) while being nearly invisible on ChatGPT (where third-party directories it neglects drive most citations). Our AI tool adoption rates report tracks how marketing teams are shifting measurement approaches in response to the growing AI platform landscape.

What Does a Strong Share of Model Score Look Like?

LLM Pulse benchmark data segments Share of Model performance into four tiers: category leaders typically hold 40 to 70% SOM, top-three challengers capture 20 to 35%, top-ten players hold 10 to 20%, and new entrants sit between 2 and 10%. These benchmarks give analytics teams a baseline for evaluating competitive position.

The benchmark tiers in context:

• Category leader (40-70% SOM): The dominant brand that AI platforms name first and most frequently. In SEO tools, this likely corresponds to a brand like Semrush or Ahrefs. Only one or two brands per category typically achieve this tier
• Top-3 challenger (20-35% SOM): Brands that appear regularly but not dominantly. AI platforms mention these brands as strong alternatives. In a competitive category, holding 25% SOM means approximately one in four AI responses includes the brand
• Top-10 player (10-20% SOM): Brands with meaningful but not commanding AI presence. These brands appear in AI responses when prompts are specific to their strengths but may be absent from general category queries
• New entrant (2-10% SOM): Brands beginning to register in AI training data and retrieval sources. At 5% SOM, a brand appears in roughly one out of every twenty AI category responses

For marketing managers setting KPIs, these benchmarks translate directly to goal-setting. A brand currently at 8% SOM targeting "top-10 player" status needs to reach 10-20% — a specific, measurable objective that can be tracked monthly. Pairing SOM tracking with conversion data (AI-referred visitors convert at 7.1% per Similarweb) lets teams model the revenue impact of moving from one tier to the next. Use the ROAS calculator to model how shifts in AI-referred traffic volumes would affect overall return on marketing investment.

For brand strategists, the steep concentration at the top (leaders holding 40-70%) confirms that AI visibility follows a power-law distribution. The gap between a 15% top-ten player and a 50% category leader is not just a difference in degree — it reflects fundamentally different levels of AI model familiarity with each brand. Closing that gap requires sustained, multi-quarter investment in the content and data signals AI platforms use to generate recommendations.

How Can Marketing Teams Improve Their Share of Model?

Marketing teams can improve Share of Model by restructuring existing content for AI citation patterns — tables earn 2.5x more citations than prose, cited URLs average 17x more list sections than uncited ones, and adding schema markup yields approximately 13% higher citation odds. The optimization playbook differs substantially from traditional SEO.

Published research identifies specific content characteristics that correlate with higher AI citation rates:

Content structure signals:

• Tables outperform prose by 2.5x for AI citations. AI models parsing structured information favor clearly formatted comparison tables, specification lists, and data matrices over narrative paragraphs
• Cited URLs contain 17x more list sections than uncited URLs. Bullet-pointed lists, numbered steps, and structured frameworks make content more parseable for AI retrieval systems
• Schema markup provides approximately 13% citation odds lift. JSON-LD structured data helps AI platforms understand what a page covers, who published it, and how current the information is
• Adding statistics to content correlates with 30 to 40% AI visibility improvement. Pages containing specific data points, benchmarks, and quantified claims attract more AI citations than opinion-driven content

Content maintenance signals:

• Pages updated within 12 months are 2x more likely to retain citations. AI platforms increasingly weight content freshness, making regular updates a competitive necessity rather than a nice-to-have
• Comprehensive pieces of 2,000 or more words outperform shorter content. While traditional SEO has debated optimal content length for years, AI citation data shows a clear preference for thorough coverage. Depth signals expertise to retrieval systems

Topical authority signals:

• 40 deep interlinked pieces on one topic outperform 400 shallow pieces. AI models recognize topical clusters — a site publishing comprehensive, interconnected content on a subject earns more citations than a site publishing brief, unconnected pieces across many topics
• Reddit presence correlates with Perplexity visibility (r = +0.395, p = 0.002). Genuine participation in Reddit discussions about your product category creates citation pathways, particularly for Perplexity's retrieval system

For marketing managers building quarterly content plans, the data suggests a shift from volume-based content calendars to depth-based ones. Rather than publishing 20 thin blog posts per month across many topics, teams should consider concentrating resources on 5 to 10 comprehensive, data-rich, well-structured pieces within their core topic clusters. Every piece should include structured data markup, specific statistics with citations, and tables or lists that make key information machine-parseable.

For analytics teams measuring content performance, traditional metrics like organic traffic and time-on-page should now be supplemented with AI citation tracking. A page might generate modest organic search traffic but hold outsized influence on Share of Model because AI platforms cite it frequently in their responses. Understanding which content assets drive AI citations — separately from which assets drive search traffic — enables smarter resource allocation.

The agentic commerce trend adds additional urgency to content optimization for AI citation. As AI agents increasingly intermediate purchasing decisions, the content that AI platforms cite during product recommendation queries becomes a direct commercial asset rather than just a brand awareness play.

Which Tools Can Track Share of Model Right Now?

The Share of Model tracking landscape ranges from enterprise platforms starting with six-figure contracts to accessible tools beginning at $27 per month, plus one free option. The industry generated an estimated $100 million or more in spending on AI tracking tools in 2025 — yet 45% of marketing leaders still report an inability to accurately measure AI brand visibility.

Here is the current tool landscape sorted by accessibility:

Free:

• Seer Interactive Google Sheets Template — A free, community-built template for basic AI visibility tracking. Suitable for teams testing the concept before committing budget, though limited in automation and scale

Entry-level ($27-99/month):

• Otterly AI (from $27/month) — Lightweight AI visibility tracking designed for individual marketers and small teams. Provides basic Share of Model measurement without the complexity of enterprise platforms
• Semrush AI Visibility ($99/month add-on) — Draws from 126 million or more monthly prompts. Integrates with Semrush's broader SEO toolkit, making it a natural extension for teams already using Semrush for traditional search visibility tracking
• Profound (from $99/month) — Tracks 10 or more AI engines with a dataset of 400 million or more conversations. Profound's 6.8-million-citation dataset provided much of the platform-specific citation pattern data cited earlier in this guide

Mid-tier ($199-699/month):

• Ahrefs Brand Radar ($199-699/month) — Monitors 185 million or more monthly prompts. For teams already using Ahrefs for backlink and keyword analysis, Brand Radar adds AI visibility tracking to the existing workflow

Enterprise:

• Jellyfish Share of Model — The original trademarked platform, tracking ChatGPT, Gemini, and Llama. Beta clients include Danone and Pernod-Ricard. Enterprise pricing, designed for global brand portfolios requiring multi-market, multi-model tracking
• Evertune — The only platform currently offering AI retargeting through a Trade Desk integration. Enterprise positioning with a unique capability: not just measuring AI visibility but acting on it through programmatic advertising

For marketing managers evaluating tools, the decision hinges on existing stack. Teams already invested in Semrush or Ahrefs benefit from their respective AI visibility add-ons, which avoid the overhead of onboarding an entirely new platform. Teams without an existing SEO tool relationship might find Profound's multi-engine coverage or Otterly's low entry price more practical starting points.

For analytics teams building measurement infrastructure, the SparkToro finding about ranking inconsistency should inform tool evaluation. Any tracking tool that reports a static "ranking position" for your brand in AI responses is presenting misleading data. The meaningful metric is mention frequency across large prompt samples — evaluate whether a tool's methodology aligns with this statistical reality before committing budget. B2B SaaS discovery through AI answers has grown to 17% (up from 4% the prior year), and 89% of B2B buyers now use generative AI for vendor research — the measurement gap is closing, but the tool landscape remains fragmented.

Meanwhile, 66% of Gen Z consumers request AI brand recommendations according to YouGov research surveying 1,000 US consumers — a demographic signal that Share of Model will only grow in strategic importance as AI-native consumer cohorts gain purchasing power.

Where Can I Learn More?

• Semrush Review 2026 — How Semrush's AI Visibility add-on tracks brand mentions across 126M monthly AI prompts alongside traditional SEO metrics
• Ahrefs Review 2026 — Ahrefs Brand Radar for AI visibility monitoring paired with the platform's backlink and keyword research capabilities
• AI Tool Adoption Rates 2026 — How marketing teams are adopting AI tools across categories, including visibility tracking and analytics platforms
• Amplitude vs Mixpanel — Side-by-side comparison of analytics platforms as teams evaluate where AI visibility data fits into their measurement stack
• Google Analytics Review 2026 — Tracking AI-referred traffic and conversion rates within your existing analytics infrastructure
• What Is Agentic Commerce? — How AI agents are beginning to intermediate purchasing decisions, making Share of Model a direct commercial metric
• ROAS Calculator — Model the revenue impact of AI-referred traffic using the 7.1% conversion rate benchmark

Sources

• Similarweb — 45 billion monthly AI platform sessions (56% of search volume), 7.1% AI-referred visitor conversion rate
• Semrush — 126M prompt analysis (Jan-Apr 2026), "Universal 36" brands finding, 45% marketing leaders unable to measure AI visibility, 9% possessing tools
• SparkToro — Rand Fishkin — January 2026 study: 600 volunteers, 2,961 prompts, less than 1-in-100 chance of same brand list, 1-in-1,000 chance of same order
• Profound — 6.8M citation analysis: platform-specific citation rates (ChatGPT 48.73% directories, Gemini 52.15% brand sites, Perplexity 21.87 citations/answer, AI Overviews 59.8% brand domains, Reddit ~40% cross-platform)
• Jellyfish / The Brandtech Group — Share of Model concept origin (Jack Smyth, Tom Roach, late 2024), commercial platform, Danone and Pernod-Ricard beta clients
• Symphonic Digital — Share of Model definition: "the number of mentions of a brand by one or multiple LLMs, as a proportion of total mentions of brands in the same category"
• Digital Applied — 14% of marketers tracking AI citations vs 43% naming it core strategy
• YouGov — 66% of Gen Z request AI brand recommendations (1,000 US consumers)
• LLM Pulse — SOM benchmark tiers: leader 40-70%, top-3 challenger 20-35%, top-10 player 10-20%, new entrant 2-10%
• Ariel/Chanteclair cross-platform SOM data — Ariel 24% on Meta Llama vs less than 1% on Gemini; Chanteclair 19% on Perplexity vs 0% on Llama
• Industry estimates — $100M+ annual spending on AI tracking tools, 52% of brands never surface across AI engines, 11% domain overlap between ChatGPT and Perplexity