ARE AVMs THE SAME AS AI METHODS? A LITERATURE REVIEW

Abstract This review examines whether Automated Valuation Models (AVMs) and Artificial Intelligence (AI) are synonymous within real-estate valuation. Drawing on 37 academic and grey-literature sources (including IVS, RICS, IAAO, and recent empirical studies), the paper clarifies the conceptual and operational relationships between AVMs, CAMAs, and AI/ML methods. The results indicate that AVM is a domain-level framework for automated property valuation that has historically used statistical or rule-based methods (like hedonic regression). On the other hand, AVMs can utilise AI, a group of adaptive algorithms: Neural networks, tree ensembles, SVMs, and computer vision. Modern AVMs form a spectrum—from purely statistical to fully AI-driven systems—and AI generally improves predictive accuracy and enables use of novel, unstructured data (images, text, geospatial metrics). However, AI-enhanced AVMs raise challenges in terms of interpretability, data quality and privacy, regulatory acceptability, and professional oversight. The review concludes that AVM ≠ AI, but AI is increasingly integral to state-of-the-art AVMs; it recommends standardised terminology, comparative evaluations across market types, explainable and auditable AI methods, and hybrid human–AI workflows to balance predictive gains with transparency and accountability.