AI-assisted conversational search is rapidly displacing filter-based interfaces across the major home search portals. Redfin's deployment of conversational search produced a 47\% lift in tour requests, and Zillow launched "AI Mode" in March 2026. Recent consumer surveys indicate that a large majority of Americans now use AI tools for housing market information. Yet the evaluation frameworks practitioners apply to these products remain borrowed from general-purpose usability testing, tools designed for deterministic, filter-driven interfaces that do not capture the distinctive failure modes of AI-driven experiences. This paper proposes a four-layer quantitative evaluation framework purpose-built for AI-assisted home search: recommendation system quality, interaction efficiency, attitudinal measurement, and trust calibration. For each layer, validated instruments, production-derived benchmarks, and practitioner-ready implementation guidance are provided. A minimum viable metric set and a worked example illustrating the framework's application to a mid-sized portal are included to support immediate adoption.

翻译：暂无翻译