When a customer asks an AI about your category, does your product appear? With what arguments? Against whom? mAIr Arena measures what AI claims and displays about your product, compared with others — an observed fact, never a value judgment.

How it works

Three angles of comparison

Presence

Is your product in the arena?

From memory (model knowledge), no web search: does AI know your product, yes or no. The shortlist — often decisive in B2B.

With web search (live): is your product cited when AI answers a buying intent. Often decisive in B2C.

Arguments

What AI says for — and against

The arguments AI puts forward in favor of your product, and the ones it spreads against it.

The "against" arguments are sometimes uncomfortable. That's precisely their value: you see what your customers read before they buy.

Competitors & market

Against whom, and at what displayed price

The competing products AI cites, their arguments, the sources it draws on, and the transactional block: purchase links, presence, displayed prices, price gaps.

Red line — Arena is a mirror of what AI displays, never a judge of true or false. A captured price is an observed, frozen data point — we attest "on this date, on this query, AI displayed X", not "this product costs X".

What Arena measures

Each measurement compares your product with others, on the same queries, at the same moment.

  • Presence — Is your product cited, in memory and on the web: an observed rate, not an impression.
  • Competitors — Which comparable products AI cites against yours, and with what arguments.
  • Transactional — What AI displays: purchase links, prices, price gaps. Captured, dated, sealed.

Model knowledge and live web are always provided. The interpretation is yours: AI measures, you decide.

Why it's reliable

  • Repeated — n=20 per measurement: each rate is a fact observed over 20 queries, not a single answer. AI's instability is measured, not endured. Beyond that, our tests show that increasing the number of queries yields no more relevant result.
  • Mechanical — All scores are computed in Python from the raw answers, through an in-house process. AI answers; it never computes its own score.
  • Model knowledge and live web, separated — What AI knows from memory and what it finds on the web are measured separately, never mixed. When a customer queries an AI, nothing tells them whether the answer comes from the model's memory or a web search — yet the two don't respond to the same levers. Knowing your position on each equips you to adjust your strategy.
  • Same protocol for all — Your product and the competitors are queried on the same prompts, at the same instant. The comparison is on a level playing field.
  • Sealed — Every report is dated and sealed (HMAC). You can prove, later, what AI displayed that day — no one else can.

Arena doesn't say which product is best. It attests what AI claims and displays, on a given date, on given queries.

Who it's for

For brands that want to know how their product is presented against the competition in AI answers — and prove it.

For agencies managing their clients' product visibility. See the agency offer

Measure your product in the arena

Build your measurement, order, receive a dated and sealed report. You have nothing to compute.

Each measurement hands you a full report and a neutral prompt, usable with your LLM, your agency or your teams — the material is yours.

How it works