Share of Model: the metric that replaces share of search in the AI era

What Share of Model measures

Share of Model (SOM) measures how often and how favorably a brand is mentioned in responses generated by large language models — ChatGPT, Gemini, Claude, Perplexity, and the other AI systems that are increasingly the first point of contact between a user and information about a category.

It is not a single number. It is a composite of several dimensions:

Mention rate: what percentage of relevant prompts about your category include a mention of your brand name in the AI-generated response?

Position: when your brand is mentioned, where does it appear in the response? First in a list, mid-list, or as a secondary afterthought? Position within an AI response correlates with the authority signal the mention conveys.

Sentiment and attribute association: what does the AI say about your brand when it mentions you? Is the description accurate, differentiated, and positive? Does it reflect the positioning you have built, or a diluted or outdated version of it?

Consistency across platforms: does your brand appear consistently across ChatGPT, Gemini, Perplexity, and Copilot — or strongly on one platform and absent on others?

Share of voice relative to competitors: not just "am I mentioned?" but "how often am I mentioned relative to the other brands in my category, across the full distribution of relevant prompts?"

Taken together, these dimensions produce a picture of your brand's presence in the AI layer — the interface that sits between user intent and brand discovery for an increasingly large share of the population.

Why share of search no longer captures AI-era brand visibility

Share of search was built for a specific architecture: the user formulates a query, receives a list of results, and navigates to a website. Measuring query volume for your brand name — relative to competitors and to total category query volume — gave you a reliable proxy for brand awareness and consideration.

That architecture has three assumptions that no longer hold universally.

Assumption 1: the user formulates a query. In AI-mediated discovery, users often do not formulate a discrete search query. They ask a question in natural language, have a conversation, or delegate a task to an agent. These interactions do not generate query data in the traditional sense. Share of search cannot measure presence in a conversation.

Assumption 2: the result is a list of links. AI-generated answers are not lists. They are synthesized responses that name specific brands, describe their attributes, and make implicit or explicit recommendations. A brand mentioned favorably in a synthesized response does not necessarily generate a click — but it generates awareness, consideration, and trust in a way that appearing as a blue link on a results page does not.

Assumption 3: visibility is binary. In traditional search, you either rank for a query or you do not. There is a threshold below which you are invisible. In AI-generated responses, visibility is graduated — you can be mentioned first, mentioned with strong attribute language, mentioned briefly in a list, mentioned with a qualifier, or not mentioned at all. This gradient matters enormously for brand-building, and share of search cannot capture it.

Share of Model is designed for the new architecture. It measures presence in synthesized responses rather than in ranked lists, captures the quality and position of mentions rather than just their existence, and tracks consistency across AI platforms rather than just across search engines.

How to measure Share of Model in practice

SOM measurement is more complex than share of search measurement, but the methodology is tractable. Here is how to approach it.

Step 1 — Build your prompt library. Identify the 50 to 100 prompts that are most relevant to your brand's category. These should include definitional prompts ("what is the best [category] for [use case]?"), comparison prompts ("compare [your category] options for [specific need]"), recommendation prompts ("which [category] should I use if [specific context]?"), and attribute prompts ("which [category] is known for [specific capability]?"). The prompt library should represent the full range of questions your target users actually ask AI systems.

Step 2 — Run prompts across platforms. Submit each prompt across ChatGPT, Gemini, Perplexity, and Copilot — at minimum. For each response, record: does your brand appear? At what position? In what context and with what language? What competitors appear alongside or instead of you? Run each prompt multiple times, across different sessions, to account for response variability.

Step 3 — Score each dimension. Build a scoring framework for mention rate, position, sentiment, and competitive share across each platform. Standardize the scoring so that results are comparable over time and across platforms. This is the SOM dashboard — your ongoing view of where you stand in the AI layer.

Step 4 — Track over time. SOM is a leading indicator, not a lagging one. Changes in your SOM — up or down — typically precede changes in AI-referred traffic, branded search volume, and eventually revenue. Track it monthly. Compare against your prompt library each cycle. Identify which prompts are improving and which are declining, and use that data to diagnose what is driving the movement.

Step 5 — Benchmark against competitors. SOM is most useful as a relative measure. Your absolute mention rate matters less than your mention rate relative to the three or four competitors that your target users are actually comparing you against. Build a competitive SOM dashboard that shows your share alongside theirs, per platform and per prompt category.

The tools that measure SOM today

The tooling ecosystem for SOM measurement has developed rapidly in 2026. Several platforms now offer varying levels of SOM tracking capability.

Profound — the enterprise market leader, backed by $35M in Series B from Sequoia Capital. Tracks brand mentions across 10+ AI engines including ChatGPT (GPT-5.5), Claude, Perplexity, Google AI Overviews, Gemini, Copilot, DeepSeek, and Grok. Includes 400M+ real user prompts via Conversation Explorer and AI crawler analytics. Priced for enterprise; requires sales engagement.

Evertune — tracks mentions across all major LLMs with source attribution (connecting mentions back to the specific content that drove them), sentiment analysis, and competitive benchmarking. Purpose-built for the full SOM picture, not just citation links.

OpenLens — the first free platform for multi-platform AI brand monitoring. Covers ChatGPT, Claude, Google AI, Perplexity, and DeepSeek. No credit card required. Limited depth compared to enterprise platforms but zero cost — the right starting point for brands building their first SOM baseline.

Semrush AI Toolkit (now Adobe) — integrating AI visibility tracking into the established SEO workflow. Accessible for teams already in the Semrush ecosystem.

Ahrefs Brand Radar — overlays AI citation frequency with traditional SEO metrics, useful for teams wanting to understand the correlation between organic authority and AI mention rates.

The most important tool is whichever one produces a consistent, comparable, platform-specific view of your brand's presence in AI responses over time. Consistency of methodology matters more than feature breadth — a simple, consistent tracking system produces more actionable insight than a sophisticated system used inconsistently.

SOM as a leading indicator: what the data reveals

The most compelling case for SOM as a strategic metric comes from the predictive relationship between AI mention patterns and downstream business outcomes.

Research from Conductor (2026 AEO/GEO Benchmarks Report) identifies AI visibility as its own performance channel — distinct from organic search, paid, direct, and social. Brands that track SOM are not just measuring a new metric. They are tracking the channel that increasingly determines which brands enter consideration before a user ever performs a traditional search.

Harvard Business Review's 2026 survey of 12,000 consumers found that 58% used AI to search for products and services in 2025 — up from 25% in 2023. These users skew younger, wealthier, and more educated. They are the demographic cohort with the highest lifetime value and the lowest traditional advertising receptivity. The brands they encounter in AI responses are the brands they consider. The brands they do not encounter are effectively invisible to this cohort's discovery process.

SOM is the metric that tells you how visible you are to that cohort. Share of search tells you how visible you were to the cohort that used Google. These are increasingly different populations.

Conclusion

Share of search was the right metric for the era it measured. It captured brand presence in the dominant discovery interface of its time with reasonable precision and reasonable ease of measurement.

The dominant discovery interface has changed. For a significant and growing share of the population — and for the highest-value demographic cohorts — the discovery interface is an AI-generated answer, not a list of search results. Share of search does not measure presence in that interface. Share of Model does.

The brands that build SOM measurement infrastructure now — prompt libraries, platform-specific tracking, competitive benchmarking, time-series baselines — are building the analytical foundation for AI-era brand strategy. The brands that continue to measure only share of search are measuring their presence in a channel whose relative importance is declining, while remaining blind to a channel whose importance is compounding.

That is not a sustainable measurement strategy. And it is entirely fixable, starting today.