Understanding share of model
Every GEO campaign needs a north star metric. For AI visibility, that metric is Share of Model, and if you're not tracking it, you're flying blind.
What share of model actually measures
Share of Model (SOM) answers one question: when people ask AI models buying-intent questions in your category, how often does your brand appear in the answer?
The formula is simple:
SOM = (successful, relevant answer samples mentioning your brand / successful, relevant answer samples tested) x 100
Arithmetic example only: If eight of fifty comparable successful answer samples mention your brand, the sampled SOM is 16%. Failed and skipped checks are separate coverage information.
Why SOM is useful
SOM summarizes brand presence in a defined set of sampled answers. Read it alongside the actual answers, source evidence, successful sample counts, and the business question being tested. It is one diagnostic measure, not proof of which tactic worked or a substitute for customer outcomes.
Interpreting a sample
Any low, middle, or high label is only a shorthand for the particular prompt set. There are no universal category tiers or assured starting ranges. A high mention rate does not establish market dominance, and a zero rate does not show that a model knows nothing about the brand.
SOM vs mentioned vs recommended
Outrigger tracks three levels of brand presence in AI responses:
- Mentioned: Your brand name appears anywhere in the response. This includes passing references like "companies like X, Y, and Z."
- Recommended: The AI model specifically suggests or endorses your brand. This is stronger than a mention: the model is saying "you should check out X."
- Linked: The AI model includes a URL pointing to your website or a page about you. This is the strongest signal and most common in Perplexity responses.
SOM counts mentions (the broadest measure). But the recommendation rate and link rate are tracked separately because they tell you about the quality of your visibility, not just the quantity.
Why you track over time, not as a snapshot
A single SOM measurement is a starting point. The real value comes from trending:
- Regular tracking can show changes in comparable sampled answers.
- Longer comparisons can add context, but a calendar interval alone does not ensure sufficient evidence.
- Before/after work describes timing. It does not prove that the work caused an observed change.
AI model responses aren't static: they change as models update their training data and as live-search results shift. A sampled rate can change without new work by the brand. Review sampling, engine behavior, sources, and competitor context without assuming a cause.
The competitor dimension
SOM isn't just about your brand: it's relative. Knowing your competitor's SOM tells you:
- Who's winning in each AI model
- Which models are the biggest opportunity (if competitors dominate ChatGPT but nobody owns Perplexity, that's your opening)
- Whether the gap is growing or shrinking over time
In Outrigger
Sonar dashboard shows your current SOM front and center: the big number at the top. Below it, you'll see the model-by-model breakdown, keyword performance, and competitor comparison. Everything updates after each monitoring run.