AI brand tracking
When customers ask a model about your category, its answer is already influencing what they buy. This measures that answer systematically: brand by brand, attribute by attribute.
Interactive demo: choose a category, then a lens, to see how an AI model ranks a category's priorities.
For more decisions, the first opinion a customer encounters is becoming a model's. They ask it which mattress to buy, which CRM to choose, which running shoe fits them, and its answer becomes part of the decision process, often before a search or a review. The model now holds a working picture of your category, and that picture is influencing demand whether anyone is measuring it or not.
AI brand tracking measures that picture. I take a category and interrogate the model like you would design a research study: mapping what it believes the category is about, the jobs customers hire it for, the capabilities they judge it on, and the perceptions attached to each brand.
A quick note on scope. This is a first-pass category map, designed to reveal the shape of a market quickly. The three lenses, jobs, capabilities and perceptions, capture three important layers of meaning, but no category is reduced to three dimensions. Every market has attributes that matter operationally, attributes customers argue about, and attributes that only become visible when you look deeper. Mapping that full landscape is the research itself. The value of tracking comes from understanding what matters most, and where brands actually differ.
How to read it
Choose a category, then a lens. The bars show the model's ranking of what matters in that category. Jobs to be done are the outcomes people are trying to achieve. Capabilities are the functional criteria they judge a product or service on. Perceptions are the associations and judgements attached to brands.
Why not just ask ChatGPT
Anyone can ask a model a question and capture the answer. Turning that answer into a measurement requires a method.
- Control. The context is designed deliberately and held constant. Changes between runs need to represent real shifts, not accidental differences in wording or setup.
- Comparison. A category looks different depending on who is looking. The useful question is often not what the average customer sees, but how the picture changes between segments, occasions and priorities.
- Consensus. One answer is not a measure. Repeated runs reveal the stable patterns underneath the variation, allowing meaningful differences to be separated from noise.
Beyond the demo
The demonstration ranks the attributes. That map is the starting point, not the finish. The same attributes become the frame for a set of sharper questions:
- What matters most. Which attributes actually drive choice, not just the ones a model lists first. Importance gets separated from noise, so effort goes where it moves the decision.
- Where dissatisfaction lies. Cross importance with how well the category performs and the gaps appear: the things that matter and are being done badly. That gap is the opportunity.
- What is profitable. Not every improvement earns its keep. The attributes people will pay a premium for are marked off from the table stakes everyone has to meet.
- Where competitors lead. Scored brand by brand, the map becomes a competitive picture: who owns which ground, where a rival is exposed, and which strengths are real rather than merely claimed.
- The open ground. Attributes with high demand and no strong incumbent are where a challenger plants a flag. High importance with weak competition is whitespace worth taking.
- Substitutes and boundaries. What people reach for instead, and where the category bleeds into the ones next door, shows where demand leaks and how wide the real competitive set actually is.
Each of these is a different instrument pointed at the same map. Held together, they turn a description of a category into a decision you can act on.
One caveat matters. These readings are not the opinions of customers. They are measurements of the model's representation of a category. That makes them a source of hypotheses, not a replacement for human research. The goal is not to ask AI for the truth. It is to understand what AI believes, identify what is worth testing, and decide where deeper research should go.
The tools that read this layer are new, and the category rushing to fill the gap is mostly selling confidence without a method. That is the opening. If you want to know what the model is telling your customers about your category, and you want an answer you can stand behind in a room, that is the work I do.
Start a conversation