The rapid proliferation of digital commerce has heightened the need for adaptive, data-driven pricing mechanisms capable of responding to complex and rapidly evolving consumer behaviors. Conventional approaches—such as rule-based heuristics and econometric models—struggle to capture nonlinear market dynamics and behavioral heterogeneity inherent in online transactions. This study investigates the application of deep reinforcement learning (DRL) for dynamic pricing in e-commerce, leveraging real-world transaction datasets to evaluate its empirical effectiveness. Specifically, Deep Q-Network and Advantage Actor-Critic models are developed to optimize pricing decisions through iterative interaction with simulated market environments. Data preprocessing encompasses customer segmentation, price elasticity estimation, and temporal normalization to ensure model robustness across heterogeneous retail contexts. Experimental results reveal that DRL-based pricing strategies outperform traditional benchmarks, yielding 14–21% improvements in revenue per visitor and measurable reductions in customer churn across multiple product categories. Furthermore, the adaptive exploration–exploitation dynamics of DRL enable the models to capture nuanced behavioral signals—such as sensitivity to promotions and seasonal variations—that static methods typically overlook. Comparative analyses of forecasting accuracy, profitability, and customer retention substantiate the superior performance of DRL architectures. By demonstrating how DRL can autonomously learn and generalize pricing policies from transactional data, this research bridges the gap between algorithmic optimization and real-world consumer behavior. The findings establish DRL as a robust and scalable framework for intelligent pricing in digital marketplaces and highlight future directions for integrating explainability and fairness constraints to support transparent, ethical, and sustainable AI-driven commerce.