Inside a Brand Reading: how we read what models understand about your brand.

Every Monday morning, the brands that work with us receive a one-page document called a Brand Reading. It is the answer to a question their CMO has been asking, in some form, for the past eighteen months: what do the AI assistants actually understand about my brand this week?

The artifact looks deceptively simple: six scores, three gaps, a short narrative, a recommended action for the Comex. The work behind it is anything but. This is the methodology, as honestly as we can describe it.

Step 1: Probing the models

Every week, AYAN runs a structured set of questions against the leading public models: currently ChatGPT, Claude, Gemini, Perplexity, Mistral, Copilot, plus selected enterprise deployments. The question set is not generic. It is sized to your category, your competitive set, and your priority markets. A luxury skincare brand and a fintech challenger receive entirely different probe matrices.

The probes are designed to elicit interpretation, not just retrieval. We do not ask “what is [brand]?” Every model answers that one. We ask the questions a real customer asks at each stage of the journey: which brands should I consider, why this one over that one, what is the quality difference, is the sustainability claim real, does the price reflect the craft.

Step 2: Reading the answers, not counting them

For every probe, we capture three things. The mention itself (yes, the visibility layer). The claims the model attaches to the mention (the interpretation layer). And the sources the model retrieved to support those claims (the provenance layer).

The interpretation layer is where most of the work happens. We extract every assertion the model makes about your brand (quality, trust, sustainability, innovation, fit, value) and score it against your Brand Base, the source of truth that captures what your brand actually intends to be. A model claim that aligns with your brand truth strengthens the dimension score. A claim that drifts weakens it. A claim that contradicts it triggers a gap.

Step 3: The six dimensions

Six is not arbitrary. After eighteen months of running readings across categories, six is the smallest set that captures the meaningful variance and the largest set a CMO can hold in their head during a meeting.

Quality is what the model claims about the craft, the materials, the manufacturing. Trust is what the model claims about reliability, longevity, customer experience. Sustainability is what the model claims about environmental and social commitments. Innovation is what the model claims about how the brand pushes its category forward. Fit is what the model places you in: your category, your peer set, your customer. Value is what the model claims about the price-to-promise relationship.

Every dimension is scored 0–100, segmented per model and per market. The headline score on the Brand Reading is a weighted blend; the underlying detail is available for any dimension that moves.

Step 4: Tracing the source

Every claim the model makes is traced, as far as the model will let us, back to the public signal that fed it. A retailer PDP. An old press release. A Reddit post. A founder interview. A Wikipedia edit. This is the most expensive part of the work, and the most useful one, because a claim with a known source is a claim a brand can correct. A claim without a source is a guess.

Step 5: Writing the reading

The reading itself is written, not generated. A senior analyst takes the scores, the gaps, the source traces, and writes the one-page narrative that lands on your desk Monday morning. We could automate this. We have tried. The trade-off is not worth it. A CMO can tell within two paragraphs whether they are reading something written by a person who thought about their brand or by a template. The first kind gets opened next week. The second kind does not.

Step 6: Closing the loop

The Brand Reading is not the end of the workflow. Every gap it names becomes an owned action, routed to the team that controls the surface where the fix needs to land. Web team, retail team, content team, sales team, care team. We track each fix through to landing, and the next Brand Reading verifies whether the model interpretation moved.

That is the methodology. Probes, claims, provenance, scoring, writing, closure. The artifact looks simple because the work behind it is precise.