Assess whether your category is suitable for AI Visibility measurement.
Start with a free category-fit assessment. We will confirm whether a Snapshot, an audit or work on another foundation is the right next step.
TL;DR
- Share of Model is the brand's share in the responses of AI models to a given set of prompts - how many times and with what quality the model recommends your brand compared to the competition. A typical increase of 25-35% compared to the control 50 prompts is achieved through cyclical GEO on-page activities (schema and citation). Limitation: this share depends on the specific platform - ChatGPT treats sources differently than Perplexity, so measurement always requires contextual rigor. Related concepts: 'LLM market mindshare' or 'LLM visibility factor'.
- The term has practical significance: it affects AI visibility strategy, content architecture and how Share of Model is measured.
- It is best to use it together with a specific implementation example and a list of constraints.
Related queries and intent
Who is this resource for?
- Business owners and marketers planning AI Visibility work
- Content and SEO specialists building service pages
- Product teams and web developers responsible for structured data and rendering
Share of Model (SOM) as a metric of the AI era
In traditional SEO, the key metric was visibility (Visibility Index), based on positions in Google. In the world of AI, where the user often does not click on links but reads the generated response, it matters whether your brand 'exists' in the mind of the model at all. Share of Model measures precisely this parameter.
SOM is the percentage of your brand's recommendations in the total pool of answers to a given set of prompts. If ChatGPT mentions your company 30 times out of 100 questions about the 'best logistics software', your SOM is 30%.
AiVisible measurement methodology
SOM measurement cannot be incidental. To be reliable, it must be based on a 'frozen prompt set' - a set of frozen queries that we repeat cyclically. At AiVisible we use an advanced Quality-Adjusted SOM formula:
- Quantitative SOM: Simple percentage of brand mentions in responses.
- Sentiment & Context: Is the brand recommended (recommendation) or only mentioned (mention)?
- Citation Quality: Does the AI link to your site as evidence?
- Competitive Gap: How does your SOM compare to the category leader's SOM?
Prompt segmentation: Not just your brand
The biggest mistake is to measure SOM only for queries about your company name (Brand SOM). The real value lies in inquiries where the customer doesn't know you yet. We divide the measurement into four segments:
- Direct Brand Prompts: 'What do you think about company X?' - here SOM should be almost 100%.
- Problem-Driven Prompts: 'How to solve the problem with [Y]?' - this is where we build expert authority.
- Category-Driven Prompts: 'Recommend the best agency [industry]' - this is where you win the market.
- Comparison Prompts: 'Company X or Company Z - what to choose?' - what matters here is the precision of the offer data.
Why is SOM more stable than the position in Google?
Google algorithms change every day (core updates), which causes sudden jumps in positions. AI models (especially those without access to live search) change their 'worldview' less frequently - usually with large updates to the model weights or the RAG database. This makes SOM a more strategic metric, showing a brand's lasting footprint in the digital collective knowledge.
Investing in Share of Model means building brand capital that is resistant to minor technical fluctuations of search engines.
Frequently asked questions
How to calculate Share of Model for your company?+
Design a control set of questions that represent real customer intentions, select models, and predetermine repetitions and coding rules. Share of Model can be calculated as the share of responses containing the brand in the entire set of observations. Always publish the result along with the measurement range and date.
How does Share of Model differ from Share of Voice?+
Share of Voice measures visibility in traditional media and search engines (links, ads, mentions). Share of Model measures share of AI system recommendations - whether ChatGPT, Perplexity or Gemini recommends your brand for specific customer queries. You can have high SOV and zero SOM if your content is not consumed by AI models.
What Share of Model is good for a B2B company?+
There is no single good value for B2B companies. The result depends on the design of the question set, category, model and competition. We therefore interpret it relative to an explicit baseline and comparable benchmark, with no universal success thresholds.
Can Share of Model be improved without changing the content on the site?+
Only partially. Mentions in independent publications and consistent brand profiles can change the mix of sources available to AI systems. Activities on your own website are the most controllable, but their impact must be confirmed by comparable measurement, without assuming an increase in Share of Model.
Next step
Choose an audit based on this methodology
Sources
- Article: What is AI Visibility and how it changed the rules of the game in B2B
A key article defining Share of Model in the context of the B2B market
- Service: Visibility diagnosis in AI
How do we measure Share of Model in practice during an audit?
- Methodology: A set of control prompts for measuring AI Search
Related resources
Generative Engine Optimization (GEO)
GEO does not replace SEO. It adds a layer of understandability and citability for AI systems generating responses.
MethodologyAI Visibility Audit Methodology
This methodology describes how to collect baselines, how to evaluate AI responses and how to compare the result after implementations.
MethodologyControl Question Set for AI Search Measurement
A good prompt set must reflect the real language of users, not just brand and marketing phrases.
Want to turn this framework into a plan for your domain?
Start with a free category-fit check. If measurement is worthwhile, we will recommend a Snapshot or a strategic audit.
0 PLN · Fit assessment, without a promise of a full report