August 22, 2026
Share of voice measured how loud a brand was. Share of model measures whether an AI system actually recommends it. Here is what the metric means, how it differs from share of voice, and how to calculate and improve it.
A brand can rank at the top of Google, run a strong paid campaign, and still lose a buyer entirely inside a single ChatGPT answer that never mentions it. That is the blind spot most marketing dashboards were never built to catch, and it is why a new phrase has started showing up in board decks and marketing plans throughout 2026: share of model.
Share of model in AI search measures something specific: whether a brand gets named at all inside the answers large language models actually give, how much weight that mention carries once it is there, and whether the brand comes out of it looking good, measured against every rival brand answering the same questions. It draws on the same instinct that made share of voice a trusted metric for decades, but it asks a sharper question suited to a channel where a generative engine settles on one confident answer rather than a page of ten links.
What follows breaks down what share of model actually is, the share of voice vs share of model distinction, why it is pulling attention away from website traffic as the metric that matters, how it connects to brand awareness and market perception, how it actually gets calculated, and what a company can do about the number once it has one.
When a language model answers a category question, whether that means naming the best options, comparing two vendors, or making an outright recommendation, only a handful of brands actually get named. Share of model is the read on how often a given brand is one of them, weighed against how often every rival brand earns that same kind of mention across the identical prompt set. It is a young term. The measurement conventions around it are still settling, and different tools weigh its parts differently, but the behavior it describes is already real: when someone asks ChatGPT what CRM to buy or asks Perplexity to compare two vendors, the answer settles on a small handful of names, and a brand either makes that shortlist or it does not.
The term picked up real commercial weight in December 2024, when Jellyfish, part of The Brandtech Group, launched a dedicated Share of Model platform built specifically to track how large language models perceive brands. Danone and the Pernod Ricard owned Chivas Brothers were part of the early beta group, and the work has since been covered by Harvard Business Review and MIT Technology Review. That is not the whole story of the term today. Plenty of agencies and platforms now use share of model as a general description of this kind of measurement rather than a name tied to one product, but the origin is worth knowing, since it explains why the concept arrived with more specificity than most marketing buzzwords do.
AI share of voice is the older, more established idea here, and it is fundamentally a counting exercise: add up a brand's citations and mentions across the group of prompts being tracked, divide by the combined total every brand earns across that same group, and the result is a percentage. It is a fair, useful number, and it is the foundation share of model builds on. The share of voice vs share of model distinction comes down to what each one actually asks. Share of voice asks how often a brand showed up at all. Share of model asks a sharper question about what happened once it did, since a generative answer rarely reads like an open list the way a page of search results does. It usually settles on one clear recommendation and treats everything else as a runner up, so simply being present is not automatically the same as winning the moment.
Most of the emerging thinking on share of model converges on the same idea: it is not one number so much as three signals layered together, whether a brand appears, how much weight that appearance carries, and how it reads once it is there.
Mentions. The base layer, and the one share of voice already measures well: how often a brand's name appears at all across the category prompts a team is tracking, run against ChatGPT, Gemini, and Perplexity, whether the AI links out to a source or simply names the brand from memory without citing anything.
Position. Not a ranking in the search sense, since a generative answer is prose, not a numbered list, and a literal position number is hard to track reliably for that reason. What can be judged is prominence: is a brand the one recommendation the model leads with, one of several alternatives mentioned in passing, or the answer to a direct, branded question rather than a category one. Being named first, or being the only brand the model commits to, carries more weight than being folded into a list of five.
Sentiment. How the brand is framed once it is named. A mention wrapped in a caveat, a comparison that quietly favors a competitor, or a description that gets a feature or a price wrong is a very different outcome from a clean, confident recommendation, even though a simple mention count would treat both the same way.
Weighing these three together is what separates share of model from a raw tally. Two brands can post an identical mention rate and still sit in completely different competitive positions once position and sentiment are factored in.
Because a growing share of buyers never generate a session at all. OpenAI's own reporting put ChatGPT past 900 million weekly active users earlier in 2026, and several 2026 surveys now put the share of consumers who start certain searches with an AI assistant rather than a traditional search engine at around a third. A meaningful chunk of category research now happens entirely inside a chat window, and a brand can be recommended, or quietly passed over, without a single one of those moments ever touching an analytics dashboard.
That is what makes AI search visibility a different kind of blind spot than a slow quarter of organic traffic. A dip in sessions at least shows up somewhere. A missed mention inside an AI answer shows up nowhere, since there was never a page view to lose in the first place. Tracking AI brand visibility, not just organic sessions, is what actually shows whether that shift is working for or against a company. Our guide on the zero interface economy reshaping brand discovery covers this shift in more depth, but the short version is that traffic was always a proxy for whether a brand got found and trusted, not the goal itself, and that proxy is breaking down faster than most reporting stacks have caught up with. Share of model steps into exactly that gap: a way to see competitive standing in a channel where clicks were never going to be the evidence.
Recommended Tool
Don't guess whether ChatGPT, Claude, Gemini, or Perplexity can access your content. Analyze your site in seconds.
Check AI Visibility Try Prompt Finder
No signup required • Instant results
The two feed each other, even though one does not simply become the other. Our piece on the difference between brand awareness and AI visibility covers that distinction directly, so this is worth taking a step further: awareness is the raw material a model eventually draws from. The press coverage, the reviews, the comparison content, the forum threads, all the ordinary output of a brand being known, is exactly the corpus a language model consolidates when it decides what to say about that brand and how favorably to say it. Brand awareness in AI search, in other words, is not really a separate thing from share of model. It is the upstream input that share of model eventually reflects back.
What makes share of model genuinely new is that it turns market perception into something closer to observable, and it does it faster than traditional brand tracking ever could. A perception survey takes weeks to field and reflects a lagging view of sentiment. Position and sentiment inside AI answers move as the underlying web moves, sometimes within days of new coverage appearing. A company is not just checking whether it gets named when it tracks share of model. It is watching a live, if imperfect, read of how the wider internet currently talks about it, filtered through a model that has read far more of that conversation than any single research panel ever will.
There is no single agreed formula yet, and any guide that claims otherwise is getting ahead of where the practice actually stands. What most approaches share is a rough sequence. Start with a defined set of category prompts drawn from how buyers genuinely phrase these questions, and run them against every tracked engine, typically ChatGPT, Gemini, and Perplexity, since those three account for the large majority of usable AI search volume today. AI Overviews and AI Mode live inside Google Search rather than as a separate queryable engine, so Gemini tracking is usually the closest available proxy for both.
A simple version of the math: run 40 category prompts across those three engines for 120 total responses, and if a brand's name shows up in 30 of them, its raw mention layer sits at 25 percent, the same arithmetic AI search visibility metrics like citation rate and mention rate already use. Share of model layers on top of that: weighting appearances where the brand is the one named recommendation more heavily than a passing mention buried in a list of alternatives, then discounting appearances framed with hedges, caveats, or outright inaccuracies. The result is a more textured number than 25 percent alone, though exactly how much weight position and sentiment should carry compared with raw mentions is still something individual tools and teams are settling on their own way, which is worth knowing before treating any single share of model score as an industry standard figure.
Since the metric is built from three layers, the most useful way to improve it is to work on each layer on its own terms rather than treating it as one vague goal.
Raise mentions first. Find the specific category prompts where named competitors get cited and a brand does not, and close those gaps with content that responds directly to the exact prompt being asked. This is the fastest lever available, since it is a specific, addressable gap rather than a request for more content in general.
Earn position, not just presence. A model tends to lead with whichever source states an answer most directly and confidently. Content that commits to a clear recommendation or a specific comparison earns the primary mention far more often than content that hedges across several options at once.
Protect sentiment through consistency. A model hedges on a brand it cannot confidently verify, and inconsistent facts across a brand's own site and independent sources are exactly what produce that hesitation. Our guide on trustworthiness and entity clarity in AI search goes deeper into fixing this specific piece.
Track it on a real schedule. A single check is a snapshot, not a signal. Share of model only becomes useful once it is measured consistently and read as a trend, the same way any of the other AI visibility metrics a brand cares about need to be.
Competitor Scan Live preview
See which brands AI recommends ahead of you, and why.
Winning brand visibility in AI search takes the same patient, structural work visibility has always taken everywhere else, just aimed at a channel most companies are not measuring yet. Treat mentions, position, and sentiment as three separate problems, and share of model stops being an abstract score and starts being a genuinely workable one.
No. Raw mention count is only the first of three layers. Two brands can post identical numbers there and still end up in very different competitive spots once how prominently they are placed and how favorably they are described get factored in.
Alongside it, at least for now. The terminology genuinely has not settled industry wide, some treat the two as interchangeable, others treat share of model as a refinement layered on top of the older share of voice math. In practice, most teams are better served tracking both rather than picking one and dropping the other.
ChatGPT, Gemini, and Perplexity are the three that can actually be queried directly and tracked on a defined prompt set. AI Overviews and AI Mode are not separately queryable engines, since both live inside Google Search itself, so Gemini tracking is generally treated as the closest available proxy for what happens there too.
Sentiment matters as much as the mention itself. An enthusiastic, confident recommendation and a hedged mention buried behind a caveat both technically count as an appearance, but the two land very differently, and treating them as identical hides exactly the kind of gap a company would want to know about.
Often enough that one reading never gets mistaken for a trend. One pass only captures a single day's worth of answers, and AI systems do not repeat themselves exactly from run to run, so the real value shows up once the same prompt set is tracked consistently and compared over time rather than glanced at once and set aside.
What Is Share of Model and How Does It Differ from Traditional Share of Voice?
The Three Core Components of Share of Model: Mentions, Position, and Sentiment
Why is this metric suddenly so important compared to traditional website traffic?
How Does Share of Model Relate to Brand Awareness and Market Perception?
How Is Share of Model Calculated in Practice?
What Practical Steps Can a Company Take to Improve Its Share of Model?
Frequently Asked Questions
Summarize with ChatGPT Summarize with Claude Summarize with Perplexity
S
Satvik Mishra
Co Founder of Verseodin
Satvik Mishra is the Co Founder of Verseodin, an AI visibility platform that tracks brand citations across ChatGPT, Gemini, Claude, and Perplexity. He writes about generative engine optimization strategy and what actually works for brands trying to earn visibility in AI powered search.
Newsletter
Get the latest AI Visibility insights, GEO research, product updates, and SEO strategies delivered straight to your inbox.
Continue improving your AI visibility.
Discover what users are asking AI about your industry and uncover prompt opportunities.
Open Prompt Finder
Analyze whether AI search engines can discover and cite your website.
Run Free Report
Tutorials 12 min read
Ten criteria for judging an AEO insights platform, from prompt tracking and brand mention analysis to bot data, citation sources, share of voice, and platform coverage, plus a way to choose.
Read article
Tutorials 11 min read
A practical guide to building and managing an AI search visibility strategy: audit the foundation, optimize high intent pages, track four KPI groups, analyze citation gaps, and run an ongoing optimization cycle.
Read article
Tutorials 12 min read
A mention count cannot explain ChatGPT visibility. Here is what an AEO tool should record on every run: the prompt, the full answer, competitors, products, citations, and history, and how to read them together.
Read article
Want more AI SEO insights?
Monthly research on AI search, GEO and citation trends. No noise.
[ partner program ]
Join our partner network, whether you're an agency, consultant, or reseller. Leave your email and we'll reach out with details.
Revenue share
Earn a cut of every referral you bring on, renewing month over month.
Co-marketing
Joint webinars, case studies, and content with the Verseodin team.
Priority support
Direct line to our team plus early access to new features.
From Share of Voice to Share of Model: Rethinking Brand Visibility in AI Search | VerseOdin