Abstract
Purpose: As artificial intelligence (AI) becomes integral to e-commerce, the opacity of 'black box' algorithms presents a significant challenge to consumer trust, a critical antecedent to purchase behavior. This study investigates the impact of Explainable AI (XAI) in product recommendation systems on consumer trust and subsequent purchase intent. We aim to quantify the benefits of algorithmic transparency and compare the efficacy of different explanation styles. Methods: We conducted a between-subjects online experiment with 1,248 participants. Users were randomly assigned to one of three conditions in a simulated e-commerce environment: a control group with AI recommendations lacking explanations, a treatment group with feature-based explanations, and a second treatment group with user-based (collaborative filtering) explanations. We collected data on perceived algorithmic transparency, consumer trust, and purchase intent using validated multi-item scales. The data were analyzed using Analysis of Variance (ANOVA) and Structural Equation Modeling (SEM). Results: The findings demonstrate that providing explanations for AI recommendations significantly increases perceived transparency (F(2, 1245) = 188.7, p < .001) and consumer trust (F(2, 1245) = 115.4, p < .001) compared to the control condition. User-based explanations were found to be marginally more effective in fostering trust than feature-based explanations. The SEM analysis confirmed our mediation model, showing that consumer trust is a key mediator in the relationship between the presence of XAI and increased purchase intent (indirect effect β = 0.29, p < .001). Conclusions: Implementing XAI features in e-commerce platforms is a potent strategy for mitigating consumer skepticism and enhancing commercial outcomes. The results provide strong empirical support for the business case of investing in transparent AI, offering actionable insights for managers on designing more trustworthy and effective AI-driven customer experiences. User-centric explanations that leverage social proof appear to be particularly beneficial.