AI systems are increasingly mediating competitive purchase decisions. Recent surveys indicate that 58% of buyers use AI assistants to choose between competing brands, shifting AI outputs from informational responses to decision-stage resolution mechanisms. This paper introduces AI Recommendation Intelligence (ARI), a measurement-first discipline designed to evaluate competitive outcomes within multi-turn AI decision journeys. Unlike citation tracking or prompt optimization approaches, ARI focuses on outcome-based metrics including final recommendation win rate, conversational survival rate, competitive displacement mapping, cross-model divergence, temporal stability across model updates, and full transcript preservation. Drawing on 500+ structured inspection runs across banking, travel, automotive, enterprise SaaS, food safety, and retail categories, this paper documents three consistent structural properties of AI-mediated decision systems: Cross-model factual divergence Multi-turn outcome drift Concentrated competitive displacement These findings suggest that optimization efforts conducted without baseline measurement eliminate the control condition necessary for causal attribution and governance reconstruction. AI Recommendation Intelligence is proposed as the foundational measurement layer for both competitive strategy (AIVO Edge) and regulatory-grade evidentiary monitoring (AIVO Evidentia). As AI systems increasingly resolve market choices, selection replaces visibility as the unit of analysis. Measurement therefore becomes infrastructural rather than tactical.