The AI‑driven evolution of dynamic pricing: a semi‑systematic review and novel hierarchical classification

Dynamic pricing (DP) has become a central strategy for revenue management in the data-driven economy, supported by rapid advances in machine learning, reinforcement learning, and algorithmic optimization. Despite numerous surveys, existing reviews typically focus on specific industries or isolated methodological streams, leaving a gap for a unified, method-first synthesis. To address this, we conduct a semi-systematic review of 49 influential studies (2014–2024), identified through structured searches across major databases using predefined Boolean keywords and screening criteria. Based on this corpus, we propose a novel hierarchical taxonomy that organizes modern DP research into three pillars: (1) demand learning and estimation, (2) optimization and pricing decision techniques, and (3) strategic interaction and market response frameworks. This classification integrates methodological, behavioral, and algorithmic perspectives into a structure reflecting both theoretical foundations and practical deployment. Our analysis highlights a shift from traditional econometric models toward AI-driven approaches, including deep learning and contextual bandits, alongside emerging concerns around fairness and computational efficiency. We also synthesize DP applications across energy systems, cloud services, and mobility platforms. Finally, we outline research gaps and future directions, emphasizing opportunities in meta-learning, causal inference, hybrid optimization, and quantum-enhanced pricing.