AI Consideration Measurement Methodology, Version 1.0

Five measurements dominate GEO and AI-visibility reporting: presence, prominence, competitive share, portrayal and citation support. All five are read from a single AI response to a single opening prompt, which makes them one shared measurement event rather than five independent ones, and makes the stochasticity of that response the central measurement problem.This methodology specifies how those five metrics must be constructed to support month-on-month board reporting. It requires a declared and versioned measurement population; a frozen benchmark set kept separate from discovery; at least three exact short-window replicates per measurement cell; cell-first aggregation; clustering that respects both within-cell and common-response dependence; a measured instrument noise floor; Holm multiplicity control across the five metrics; and a paired month-on-month design in which movement is reported only where it clears both the adjusted paired test and the metric-specific reproducibility floor.Version 1.0 includes a worked calibration demonstrating that these precision and reproducibility requirements can be met in practice, and states which of the five metrics are not yet calibrated and therefore may not carry a month-on-month board status. A coding guide (Annex A) forms part of the specification.The methodology is published for the industry to use, test and challenge. The authors declare their commercial interest and do not claim the position of neutral arbiter.