What is AI brand perception?

AI brand perception is the picture of a brand that AI systems form and repeat when they answer questions about its category: whether they name it at all, how they describe it, whether they get the facts right, and which brands they place it beside. It is not what people think of a brand. It is what the machines say about it — and it can be read, measured and tracked like any other signal.

Marketers already track GEO and AEO — mentions, share of voice, sentiment — because those are what show up in AI answers. Underneath is AI Brand Understanding (AIBU): how AI models understand a brand in the first place. That understanding is measured as an AI Brand Rating: in Pitot Group's work, the published Pitot Index® score.

When someone asks an AI assistant for the best hotel in a city or a reliable family car, the answer names a handful of brands and leaves the rest out. The brands it names, the order it puts them in and the words it uses are what you see of AI brand perception. Underneath sits a reading of the brand that the model has assembled from what it has learned in training and, for some systems, from what they find on the live web.

Why it matters now

AI systems are moving from answering questions to assisting with, and potentially acting within, purchasing and decision-making. The way a model reads a brand is starting to sit between that brand and the people choosing it.

That reading also moves on its own. AI systems change as they are retrained, updated and re-grounded on the live web, so a brand's reading can shift without the brand doing anything at all. A brand that has never looked at how AI describes it has, in effect, an unmonitored channel speaking on its behalf.

What AI brand perception is made of

It is not one number. It is four questions, answered together:

• Does AI name you? When someone asks an open question in your category, are you in the answer — and where?
• Does it speak well of you? When it describes you, is the framing in your favour? When asked to choose, are you the answer or the alternative?
• Does it get you right? Of the facts that matter about you, how many does it state correctly?
• Does it place you in the right context? Are you put in the right category, with the right peers?

ARES-C, the framework behind Pitot Group's measurement, reads these through five dimensions:

• Authority — whether the brand is read as one that sets the terms of its industry, or one that follows them.
• Reputation — how the brand is regarded for quality and trustworthiness, and how praise and criticism balance.
• Expertise Visibility — whether real, demonstrated capability is legible to a machine, or only marketing presence is.
• Signal Consistency — how far the models agree, derived from the dispersion between them rather than asked directly.
• Context — where the brand appears, how widely, and what it is associated with beyond its own category.

ARES-C is the instrument behind the Pitot Index score — the AI Brand Rating that measures AIBU.

Context is easy to underestimate. The same brand can be read very differently depending on the category it is being read in. In the Pitot Index for cycle 2026-W41, Chanel scored 83.7 in Luxury Fashion and 74.1 in Beauty. Same name, same week, two different readings.

Perception is not the same as visibility

Being visible to AI and being recommended by AI are not the same thing. A brand can appear often in AI-generated answers and still be evaluated less favourably than its competitors. Counting mentions tells you whether a brand turns up. It does not tell you how the model rates it once it has.

That is the practical link to GEO and AEO. Visibility work is aimed at what appears in answers; how the models understand the brand is a separate question. AIBU is that underlying reading; the Pitot Index score is the AI Brand Rating that measures it.

Nor is AI brand perception the same as human brand perception. A model has none of a customer's loyalty or nostalgia. When the way AI reads a brand and the way people see it disagree, that distance is the information: it is often the first sign of a gap between a brand's self-image and how machines are learning to describe it.

No single model speaks for AI

Different models are trained on different data, by different organisations, under different policies and languages, and they can rate the same brand differently. A reading from one model tells you what that model says. It does not tell you whether the assessment is shared.

That is why the Pitot Index puts one standardised, unprompted test to a set of pinned AI models — currently nine, from Western, Chinese and Arabic-first developers — every week. The models are never told which brands to discuss. We check model versions every cycle and log any change we find, with the date we found it. The result shows where models agree and where they diverge. Model origin is not treated as proof of bias; differences are measured and reported where the evidence supports them.

Unprompted and prompted perception

There are two ways to ask the question, and they give different answers.

Unprompted perception is what AI brings up on its own. The Pitot Index works this way: it records which brands the models bring up in each industry, and how they rate them.

Prompted perception is what AI says when a brand is named and asked about directly. That is how a commissioned assessment of a single brand, such as Pitot Group's Audit, works.

The same brand can score very differently under the two. That is the method, not an error, and the two are never compared.

What a brand can do about it

The first step is to read it. The published Pitot Index shows how AI currently rates the top ten brands in each of fourteen industries, week by week, free to read. For a single brand, an Audit sets out what AI says, where it is wrong and where the gap begins.

A thermometer without treatment only tells you something's wrong. Treatment without a diagnosis has to guess at the cause. The Pitot Index is the diagnosis; GEO and AEO are the treatment.

What no one can honestly offer is a guaranteed place in an AI answer. The models are not controlled by the brands they describe. What a brand can control is the record they read: a consistent, independently corroborated account of what it is and what it does, maintained over time.

AI brand perception is being formed whether a brand looks at it or not. The choice is whether to read it.