What is Share of Model, and why does it matter more than your Google ranking?

For years, the question that defined a business's online visibility was a single one: what position do you rank in on Google? That question is still valid today — but it's no longer the only one that matters. There's a new metric that measures something different: how many times an AI mentions your brand when someone asks it about your industry. That's Share of Model.

The paper that formalized the concept of GEO was published by Princeton researchers in 2024 and presented at KDD, one of the world's most important data mining conferences. Its central conclusion: generative engines (ChatGPT, Claude, Gemini, Perplexity) operate on a completely different logic than traditional search engines, and conventional SEO techniques don't transfer directly to this new channel.

What Share of Model actually measures

Share of Model is the percentage of times an AI mentions your brand, product, or site when answering questions related to your category. If someone asks ChatGPT "what company sells GPS trackers in Argentina?" and the AI mentions your business in 7 out of 10 answers, your Share of Model for that query is 70%.

It's a citation-frequency metric — and like any frequency metric, it depends on how many relevant queries are run, on which AI model, and with what exact phrasing. There's no universal number: each niche, each market, and each model has its own distribution.

Princeton's work starts from a simple observation: whoever creates content has no control over when or how a generative engine displays it, because the process is opaque and changes fast. What is controllable is the form of the content. They built their benchmark on that, and there they measured that certain changes in form —not in volume or backlinks— move visibility consistently.

Why it isn't the same as your Google ranking

SEO and GEO measure different things and follow different logics. They aren't competitors — they're complementary. But confusing them leads to bad decisions.

A site can have excellent SEO — first position on Google for its most important keywords — and at the same time have a Share of Model close to zero. That means that when a user asks an AI about that niche, the site simply doesn't exist for the model. AIs don't track Google rankings: they track trustworthy, structured, citable entities.

Why Share of Model is going to matter more and more

Search behavior is changing. A growing share of queries that used to go to Google now go directly to ChatGPT, Claude, or Perplexity. The question "which GPS tracker should I buy?" is becoming "ChatGPT, which GPS tracker do you recommend for a vehicle in Argentina?" The channel through which customers arrive is shifting.

For businesses that depend on new customers finding them online — which is most businesses — ignoring Share of Model today is the equivalent of having ignored SEO in 2005. It isn't urgent yet for every niche, but the window to build presence before the channel gets saturated is closing.

A real case: how it's measured in practice

How to improve your Share of Model

Share of Model can't be bought with advertising or achieved with content volume. It's built with structure. According to Aggarwal et al., the factors with the biggest impact are:

Structured markup (Schema.org): tells the AI who you are, what you do, and what category you operate in. Without it, the model processes you as anonymous text.

Verifiable authority: visible author or founder name, with links to external profiles that confirm identity. AIs prioritize sources with identifiable people.

Concrete, citable data: statistics, metrics, percentages, dates. Vague content doesn't get cited — content with numbers does.

Topical specificity: the more focused a site is on a specific niche, the more likely an AI is to consider it an authoritative reference in that niche. A multi-category site has low Share of Model across every topic. A specialist has high Share of Model in theirs.