Updated on 19 November 2026, by João Leite, Founder of murmur.

Share of model is how AI models perceive your brand and connect it to a category. In the 2025 article by Jack Smyth and John Dawson in the Journal of Brand Strategy, that perception is read through three metrics: mention rate, average position and category associations. Monitoring tools measure the same idea under other names, and Semrush, which sells an SEO tool, has reports that show how ChatGPT, Google’s AI Mode, Perplexity and Gemini perceive the brand. The term does not have a single formula, and that is why every share of model number is read with the formula of whoever calculated it.

  • Share of model is how AI models perceive the brand. In the 2025 article by Smyth and Dawson, it is read through mention rate, average position and category association.
  • The term does not have a single formula. The article combines three metrics, and the tools give the same idea other names.
  • Share of voice is another calculation, the one for presence. Presence is how many answers the brand appears in, divided by the total number of answers.
  • Position in the answer varies a lot. In a SparkToro study, the same list of brands almost never repeated across two answers, and that is why average position is not read on its own.

Share of model is how AI models perceive your brand

Share of model is how AI models perceive your brand: whether they talk about it, where in the answer, and connected to which category. Large language models (LLMs) are the systems that write the answers from ChatGPT, Claude and the other AI models. When someone asks which company to hire, the model chooses which brands to cite and how to describe them, and share of model is the name Jack Smyth and John Dawson give to this picture, in the 2025 Journal of Brand Strategy.

The article came out in volume 14, issue 4, of the journal and describes the Share of Model platform that the two used to study brands. The article summary says that it analyzes “how LLMs perceive brands, construct category narratives and determine visibility through new metrics such as mention rate, average position and category associations”. In other words: how models perceive brands, build the narrative of each category and decide who stays visible.

In June 2025, Harvard Business Review published an article by David Dubois, associate professor at INSEAD, and coauthors, whose summary says that consumers increasingly turn to language models, and not to search engines, to answer their questions about products and services.

What share of model counts: mention, average position and category association

Share of model, in the article by Smyth and Dawson, counts three things about the brand in the models’ answers. Each name says what the metric looks at:

  1. The mention rate looks at whether the brand is cited when the question is about its category.
  2. The average position looks at where the brand appears when the answer lists several companies.
  3. The category associations look at what type of product or service the model connects the brand name to.

The three together give a broader picture than a single number. The brand can be cited often and connected to the wrong category, or appear little and always at the top when it does appear. The picture also depends on which questions go into the calculation, and that is why two share of model numbers are only compared when they come from the same questions and the same calculation.

Share of model and share of voice: what each one counts

Share of model and share of voice are different calculations: the first combines several metrics about brand perception, and the second measures the brand’s presence. Presence is how many answers the brand appears in, divided by the total number of answers. The share of mentions is the other calculation that also goes by the name share of voice, and the share of voice guide shows both, with the calculation and an example.

what is comparedshare of modelpresence (share of voice)
what it countshow AI models perceive the brand and connect it to a categoryhow many answers the brand appears in, divided by the total number of answers
how many metrics it combinesthree, in the article by Smyth and Dawson: mention rate, average position and category associationone
how the result comes outa picture of the brand, with a reading for each metrica percentage
single formulano: it depends on who calculates ityes, for presence; the share of mentions is another calculation that also goes by this name
what is left outdepends on the metrics each source combinesthe brand’s position in the answer, the tone and the category the AI connects it to

Presence is the simplest part to check: the brand is or is not in the answer. Share of model calls for more readings of each answer, and each source chooses which ones to do.

Category association: when the AI connects the brand name to what it sells

Category association is the connection the AI model makes between the brand name and what it sells. Many buyer questions mention the category, not the name: they ask which system to use, which studio to hire, which insurance to choose. A brand the AI does not connect to the category stays out of those answers, even when it is known for another reason.

The connection can come from what the model learned in training or from what it finds when it searches the internet at the time of answering. When each AI searches and when it answers from what it learned is in the guide on what comes from training and what comes from search. In both cases, what the AI has available to connect the brand to the category is what is written about it, on the company’s website and outside it.

Share of model in the tools: brand perception in each AI

Monitoring tools measure brand perception AI by AI, with their own names. Semrush, which sells an SEO tool, describes on the help page for brand performance reports that they are meant to show how ChatGPT, Google’s AI Mode, Perplexity and Gemini perceive the brand. AI Mode is Google’s search in a conversational format with the AI, instead of a list of links.

Each dashboard applies its own scale to these calculations, and the scores do not transfer from one tool to another. What each visibility score counts is in the guide on the visibility scores of each tool. The tone of the answers, which the tools call sentiment, has its own guide, on how to know whether the AI speaks well or badly about your brand.

How to track the brand’s share of model: the same questions, in each AI, over time

To track the brand’s share of model, ask the same buyer questions in each AI, repeated over time, and read the mention, the position and the category in each answer. The step-by-step for this reading is in the presence measurement guide.

Repetition counts because the answer varies. In a SparkToro study from November and December 2025, 600 volunteers ran 2,961 rounds of questions, across ChatGPT, Claude and Google’s AI. When two answers from ChatGPT or Google’s AI were compared, the list of brands came out the same less than 1 time in every 100. Fishkin, from SparkToro, started the study skeptical of this type of measurement and ended up writing, in the research report, that “visibility % across dozens to hundreds of prompts run multiple times is a reasonable metric”. The percentage of answers with the brand, across many repeated questions, holds up; the order of the list, from one answer to the next, changes a lot.

At murmur, the calculation is presence on the same questions, AI by AI: the Visio case study

murmur, a GEO agency in Brazil that places brands in the answers from ChatGPT, Claude, Google AIO and other AI agents, looks at presence, whatever name the metric goes by. Presence is how many answers the brand appears in, divided by the total number of answers. Every week, the calculation comes from the same set of questions, run in ChatGPT, Claude, Perplexity and Google AIO, the AI summary that opens many Google searches. In the dashboard, which comes with the service and is not sold on its own, the brand’s presence sits next to that of competitors.

In the Visio case study, a company that makes operations software for retail, franchises and food service, the association with the category shows up in practice. The work consisted of getting AI models to associate the name Visio with retail operations, precisely the link between brand and category that Smyth and Dawson talk about. With the same questions, between June and July 2026, Visio’s presence went from 0% to 45% of answers in Google AIO; in ChatGPT, from 2% to 22%; in Claude, from 1% to 7%. They unlocked the questions of people who did not yet know the brand, such as those looking for alternatives and competitors of other systems or asking general questions about retail management.

Frequently asked questions

Are share of model and share of voice the same thing?

No: in the article by Smyth and Dawson, share of model brings together three metrics about how AI models perceive the brand, and share of voice is presence, a single calculation. Presence is how many answers the brand appears in, divided by the total number of answers. The share of mentions also circulates under the same name, and the share of voice guide covers both.

Which metrics does share of model use?

In the 2025 article by Jack Smyth and John Dawson in the Journal of Brand Strategy, there are three: mention rate, average position and associations with the category. The term does not have a single calculation, and monitoring tools give brand perception other names. A share of model number can only be read with the calculation of whoever computed it.

Is the brand’s average position in the answer reliable?

On its own, no: the order of brands changes from one answer to another, and presence counts what repeats, which is the brand being in the answer. According to SparkToro’s research from late 2025, in ChatGPT or Google’s AI, the same list of brands came back in less than 1% of the times two answers were compared. The presence calculation has its own guide.

Which tool shows how AI models perceive the brand?

Semrush, which sells an SEO tool, has brand performance reports that show how ChatGPT, Google’s AI Mode, Perplexity and Gemini perceive the brand. The tone of those answers, which the tools call sentiment, is in the guide on how to know whether AI speaks well or badly about your brand.

Where to start

Before choosing a metric, look at the answers AI models give today in your market. With your website address, murmur’s free website diagnostic shows in less than two minutes, with no sign-up, what AI models say to people looking for what you sell, and the full report arrives by email. In murmur’s studies of each sector, you can see the brands that AI models named in each sector measured.

Read also

References