Why does the heatmap use the last 30 days?

Because a rolling thirty-day window reflects the recent state of each model without dragging in old data that no longer represents how it answers today, and because in that period there are usually enough answers per model for the percentages to be stable.

Why don't some models appear as a column?

Only models with at least one completed answer in the analyzed window are shown, since an empty column would add no information, so a model that is disabled in the project or has not answered yet does not appear on the map.

Does a competitor at 0% in a model mean that the model doesn't know it?

Not necessarily, because it only indicates that the model did not name it in the answers to the questions measured, which are those of your project, and it may well do so for other queries, so the correct reading is that in the context of your prompts that model does not recommend it.

Brand by model: where each brand appears in each AI assistant

Updated on October 4, 2026

The brand-by-model heatmap crosses the brands in your industry with the artificial intelligence models you measure and shows, in each cell, in what percentage of that model's answers the brand appears, so that at a glance you see in which assistants you lead, in which a competitor beats you and in which your brand does not exist yet.

Brand by model in a real project

In this example project the brand appears in 48% of Perplexity's answers but only in 22% of Google AI Overview's, and the most mentioned competitor never appears in ChatGPT or Claude despite being present in 13% of Perplexity's. The percentages in this example cover the whole period analyzed, whereas the dashboard shows the last 30 days by default.

	Perplexity	Gemini	ChatGPT	Claude	AI Overview
Your brand	48%	39%	29%	26%	22%
Competitor A	13%	11%	0%	0%	2%
Competitor B	8%	10%	0%	0%	2%
Competitor C	2%	5%	0%	0%	2%
Competitor D	2%	1%	0%	0%	3%

Answers analyzed per model: Perplexity: 369 · Gemini: 150 · ChatGPT: 369 · Claude: 369 · AI Overview: 246

Real data from a production project: 1,503 responses analyzed between August 6 and October 4. Brand names have been replaced with generic labels.

How do you read the heatmap?

Each row is a brand and each column a model, and the color of each cell indicates the mention rate, so dark cells mark the models in which that brand is cited often and light ones those in which it hardly appears, with a scale that distinguishes between 0%, low values of 1% to 10%, medium values of 20% to 35% and high values from 50%.

Your brand appears highlighted and, if it is not among the ten most mentioned, it is still shown in a separate row so that you always see where you stand, while brands classified as other companies and the names of artificial intelligence tools are excluded so that the map reflects only the relevant companies of your industry.

Why do the models differ so much from one another?

Each assistant builds its answers in a different way: some rely on a web search with their own preferred sources, another answers only with what it learned during training and Google AI Overview appears only for part of the queries, so it is common for the same brand to have very uneven presences across columns, as in the example, where a competitor that exceeds 10% in Perplexity and Gemini never appears in ChatGPT or Claude.

That difference is useful information in itself, because if your brand shines in the models with web search but not in the one that answers from memory, the challenge lies in the general knowledge formed about you over time, whereas if it is the other way around what is failing is your current presence in the sources those search engines consult.

What should you do about a cell at zero?

A cell at zero is not a failure but a clearly delimited opportunity, because it points to a specific model in which a brand never appears in the questions you measure, and the reasonable next step is to look in the citations section at which sources that model uses for your industry and to assess whether it makes sense to strengthen your presence in them.

It is also worth paying attention to the cells where a competitor clearly beats you, because they are the places where a change would have the most effect on your share of voice, and to check in the weekly evolution, after a few weeks, whether the cell starts to darken.

Frequently asked questions

Why does the heatmap use the last 30 days?: Because a rolling thirty-day window reflects the recent state of each model without dragging in old data that no longer represents how it answers today, and because in that period there are usually enough answers per model for the percentages to be stable.
Why don't some models appear as a column?: Only models with at least one completed answer in the analyzed window are shown, since an empty column would add no information, so a model that is disabled in the project or has not answered yet does not appear on the map.
Does a competitor at 0% in a model mean that the model doesn't know it?: Not necessarily, because it only indicates that the model did not name it in the answers to the questions measured, which are those of your project, and it may well do so for other queries, so the correct reading is that in the context of your prompts that model does not recommend it.