What is Share of Model Voice (SoMV), the key GEO metric

If SEO was measured with positions in Google, GEO is measured with Share of Model Voice. It is the metric that sums up, in a single number, how much presence your brand has inside AI answers. Understanding it is key to knowing whether your investment in generative visibility is working.

Definition

Share of Model Voice (SoMV) is the percentage of relevant prompts in your category in which an AI model mentions your brand. If out of every 100 commercial questions in your industry the models name you in 30, your SoMV is 30%. It is, in essence, your share of voice inside the "mind" of AI assistants.

How it is calculated

You first define a representative set of commercial prompts — the questions your customers would actually ask — and a market or region. Then you run them across the leading AI engines and count in how many your brand appears. SoMV can be measured per engine, per country and per prompt family, which lets you see where you are strong and where you are invisible.

Why it replaces ranking

In a generative answer there are no ten blue links: there is a synthesis with two or three brands. Either you are in it or you do not exist for that user. Traditional ranking loses meaning; what matters is how often the model chooses you as part of the answer. SoMV captures exactly that.

Difference from traditional Share of Voice

Classic Share of Voice measures your presence in media or advertising. Share of Model Voice measures your presence in the recommendations AI gives proactively, without the user clicking anything. It is a conversational market share, measured at the exact point where many purchase decisions begin today.

How to increase your SoMV

• Become legible to AI crawlers: without technical barriers, the model cannot use your content.
• Build a clear entity: consistent data and schema the model can extract and trust.
• Produce citable content: comparisons, FAQs, proprietary data and definitions that answer your industry's prompts.
• Iterate on the prompts you lose: identify where you do not appear and work those cases specifically.

Common mistakes when measuring it

Measuring few prompts, measuring once, mixing regions or ignoring the difference between engines leads to false conclusions. A useful SoMV is based on a broad set, repeated over time and segmented by engine and market. Only then does the figure reflect reality and serve to make decisions.

SoMV by engine and by market

A global SoMV hides important nuances. You might have 60% on Perplexity and 10% on Gemini, or be strong in Chile but invisible when the query comes from another country. That is why it helps to break the metric down: by engine (ChatGPT, Gemini, Perplexity, Claude, AI Mode, Copilot), by region and by prompt family.

That segmentation tells you exactly where to invest. If one engine concentrates your customers and your SoMV there is low, that is your focus. If a high-intent prompt family — the ones that signal a buying decision — ignores you, that is the content to produce. Measuring SoMV in a segmented way turns an aggregate number into a prioritized roadmap, where every percentage point gained means more chances that AI recommends you right when a customer is deciding.

The indicator that connects to the business

Raising SoMV is not a cosmetic goal: it means that when a potential customer asks AI for a recommendation, your brand is increasingly present. It is the bridge between generative visibility and the commercial pipeline, and that is why it is the heart of any serious GEO strategy. Seen this way, Share of Model Voice is not a vanity metric but the indicator that organizes the whole strategy: it sets the goal, prioritizes the work and lets you prove, with data, that the investment is paying off month after month.

In short, SoMV turns AI visibility from a vague worry into a number you can manage, report and grow — the same way you would track market share in any other channel. That is exactly why it is the starting point of any measurement effort.