Picture a hundred buyers in your category asking AI engines for a recommendation this week. The engine names brands in each answer. Count how many of those hundred answers name you, and where you land when they do, and you have a number that tells you more about your competitive position than any rankings report. That number is your share of model, and for most brands the honest count lands far lower than the founders expect. Share of model is the percentage of relevant AI answers where your brand shows up, and it is the closest thing AI search has to market share.
What is share of model? It is your slice of the total brand mentions across the set of prompts that matter in your category. If buyers ask ten core questions and the engines produce answers naming a rotating cast of brands, share of model asks what fraction of all those brand mentions are yours. A brand named in eight of ten answers has a high share of model. A brand named in one has a low one. The metric borrows the logic of share of voice from traditional marketing and points it at the place where AI answers get made, which is exactly where a rising share of buying decisions now begins.
Why share of model beats a single ranking
A ranking tells you about one query at a time. Share of model tells you about your presence across the whole category, which is what actually determines whether AI-driven buyers keep encountering you or keep missing you. Being named in one prominent answer feels good and means little if the other nine answers your buyers see skip you entirely. The value is in the aggregate, because buyers do not ask one question. They ask several, across several engines, and form their shortlist from the pattern of who keeps showing up. Share of model measures that pattern, and the pattern is what moves the market.
The aggregate view also protects you from a single lucky result. Any brand can get named once, in one answer, for one phrasing, on one engine. That is noise. A brand with real share of model gets named repeatedly, across phrasings and engines, because its presence in the sources the models read is broad and consistent enough to survive variation. When you measure share of model instead of individual mentions, you separate durable position from lucky hits, and durable position is the only kind worth building a strategy around. The metric is designed to reward the thing that compounds and ignore the thing that flickers.
There is a competitive reading here that a ranking cannot give you. Share of model is relative by construction, because your slice only has meaning against the slices your competitors hold. A 20 percent share means something completely different in a category where the leader holds 25 percent than in one where the leader holds 70 percent. Measuring share of model forces you to see the whole field at once, who dominates, who is closing, who is fading, which is the view you need to decide where to spend. Rankings show you your own position in isolation. Share of model shows you the market.
Share of model math
Here is the framework, and I keep it deliberately simple so you can actually run it. Share of model math has three inputs: your prompt set, your mention count, and the total mention count. The prompt set is the fixed list of questions that define your category, the ones buyers really ask, held constant so your measurements are comparable over time. Your mention count is how many of those answers name you. The total mention count is every brand mention across the same answers. Your share of model is your mentions divided by total mentions, expressed as a percentage. That is the core number, and it is calculable by hand from a spreadsheet of answers.
Weight it by position to make it honest. A brand named first, as the clear recommendation, is not equal to a brand named fifth in a hedged list, and a raw count treats them the same. Assign more weight to prominent mentions, less to buried ones, and your weighted share of model reflects not just whether you appear but how strongly. This is the difference between a metric that says you are present and a metric that says you are winning. The weighting scheme can be simple, first mention counts triple, mid-list counts double, trailing counts single, as long as you apply it consistently so the trend stays comparable across your monthly measurements.
Segment it by engine, because your share of model is not uniform across ChatGPT, Perplexity, Gemini, and Google’s AI answers. Each engine draws on a somewhat different source mix and composes answers differently, so you can hold a strong share on one and a weak share on another. A blended average hides that, and the gaps are where your opportunities and risks live. If you dominate one engine and vanish from another, the segmented view tells you exactly where to focus, and the blended number would have averaged that vital signal into mush. Run share of model math per engine, then look at the spread, because the spread is often more actionable than the total.
What builds share of model
Mention gravity across the sources engines read is what raises your share, because every answer is composed from those sources. When credible third-party publications, industry references, and respected sites name your brand in connection with your category, they add mass to the model’s picture of who belongs in answers about that category, and answers name the brands with the most mass. Share of model is downstream of how much of the relevant web names you. The brands with high share of model earned it by becoming heavily and consistently mentioned across the sources the engines trust, not by optimizing their own pages harder.
Consistency concentrates the mass instead of scattering it. If your brand is named across many sources but described differently, tied to different categories, or spelled inconsistently, the model struggles to build a confident association, and your share of model underperforms your raw mention volume. Uniform identity everywhere lets each mention reinforce the last, so your accumulated mentions add up into a coherent, heavy presence rather than a scattered light one. This is why two brands with similar mention counts can hold very different shares of model. The one with the consistent story converts its mentions into share. The one with the scattered story leaks it.
Category association aims the mass at the right questions. Being mentioned a lot is not enough if the mentions do not tie you clearly to the category whose prompts you want to win. Coverage and content that explicitly connect your brand to your space teach the model which answers you belong in, so your share of model rises specifically in the prompts that matter to your business rather than in random adjacent topics. The brands that build share of model deliberately do all three at once: they grow their mention volume, they keep the story consistent, and they anchor it to their category. Do that, run the share of model math monthly, and watch your slice of the answer widen against the competitors you are actually fighting.
Turning share of model into a plan
The reason to measure share of model is to act on it, and the action starts with the gap between your share and the leader’s. If the leader in your category holds 40 percent and you hold 8, that gap is your growth target, and it is concrete in a way a ranking never is. You know exactly how much of the answer space is claimable and who is claiming it. The plan is to identify where the leader’s mentions come from, the sources naming them, the prompts they own, and to build heavier, cleaner gravity than theirs in the rungs you can realistically take first. Share of model turns an abstract ambition into a measurable target with a named opponent.
Track the trend and let it grade your work. Every campaign, every earned placement, every consistency fix should show up as movement in your share of model over the following months, and if it does not, the work was not reaching the sources the models read. This is the metric that tells you whether your AI visibility spend is actually buying visibility, which is a question most brands cannot currently answer because they are not measuring the right thing. Share of model closes that loop. Measure it, weight it, segment it by engine, watch it move, and you finally have a market-share number for the place where more of your buyers now begin.