AI Discovery Radar: measurement method and ruleset for tracking adoption of machine-readable discovery and consent routes (Technical Report v1.0)

Technical report, version 1.0 (September 2026). Companion to the dataset "AI Discovery Radar — adoption measurements for machine-readable discovery and consent routes" (concept DOI 10.5281/zenodo.22178282). Adoption of machine-readable discovery and consent files for automated clients — robots.txt, llms.txt, TDMRep, agent cards, MCP manifests and others — is widely asserted and rarely counted. This report documents the method behind a monthly measurement of 34 catalogued routes on a frozen, stratified frame of 30,000 registrable domains (Tranco + Chrome UX Report; deliberately over-weighting German-, Austrian-, Swiss- and Italian-language sites). It contributes (1) a seven-state classification model with a denominator rule that separates refusal from non-adoption, (2) a two-tier, reproducible sampling design — a fixed panel for change and rotating blocks for coverage — and (3) an admission procedure for the route catalogue. Adoption here means publicly observable, syntactically valid availability of a route — deployment, not consumption. The September 2026 run illustrates the method; with two panel points under two ruleset versions, no trend is claimed. The report names its principal open question — adoption behind bot walls — and records the decision to widen the panel from 1,000 to 10,000 domains from October 2026. Measurement pitfalls (soft-404 envelopes, walls answering 404, Range-header traps, template adoption) are documented so that others can avoid them. Licence CC BY 4.0.