AI Discovery Radar — adoption measurements for machine-readable discovery and consent routes

Monthly measurements of how many websites in a frozen, stratified sample (Tranco + Chrome UX Report country lists) publish machine-readable discovery and consent files (robots.txt, llms.txt, TDMRep, agent-card.json and others). Shares over observed hosts only, Wilson 95% intervals; full ruleset with version history in RULESET.md. Data files carry run identifier and ruleset version; numbers across ruleset versions are not comparable. Each release carries the data files of its month: the fixed panel (trend series; 10,000 sources since October 2026, until December 2026 also the original 1,000-source panel as its own series) and the month's block of the wider 30,000-source frame (coverage series); every file states its measurement window and whether it is complete. After the last month of each pass through the frame, a full-frame roll-up file (quartal-YYYY-qN) counts every frame source exactly once and names the runs it is built from.