Measure AI search visibility by asking the engines the questions your real buyers ask, across several engines and several days, and scoring three things: whether your brand is named, how it is described, and whether it is recommended. A one-off prompt in one engine is a screenshot, not a measurement.
List ten questions a buyer would ask about your category. Put them to ChatGPT, Google AI Overviews and Perplexity on three separate days. Log every brand named, every source cited, and how you are described. That free exercise reveals the shape of the problem, and its limits: no repeatability, no scoring, no trend.
Scale and defensibility. A K&C report samples around 6,000 AI responses over a seven-day window, from a question set designed against the AMEC GEO Principles, scored 0 to 100 on a five-band scale with human review at six checkpoints. It also tells you what to do about it, with owner-tagged recommendations.
Start where the funnel hurts. If you don't know whether AI names you at all, start with AI See. If you're named but misdescribed, AI Think. If you're described well but missing from shortlists, AI Do.