The Role of Algorithmic Anthropomorphism, Transparency, and Fairness in Shaping Consumer Purchase Intentions in E-Commerce: Evidence from Türkiye

Artificial intelligence (AI) is increasingly being deployed in various sectors of e-commerce. Consequently, it becomes necessary to identify the impact of algorithmic design parameters on buyer behaviour. This study examines the impact of algorithmic anthropomorphism (ANT), algorithmic transparency (TRAN) and perceived algorithmic fairness (FAIR) on consumer purchase intentions (PI) in the Turkish e-commerce market. In addition, this study examines technology acceptance—operationalised through the Technology Acceptance Model (TAM)—as a boundary condition, with particular attention to the differential moderating roles of perceived ease of use (PEOU) and perceived usefulness (PU). A structured questionnaire was distributed among 384 online consumers in Türkiye via Qualtrics. A confirmatory factor analysis (CFA) established the psychometric adequacy of the measurement model (all AVE > 0.50, all CR > 0.87; HTMT < 0.85 across theoretically distinct constructs). The proposed model was tested using the PROCESS macro for sequential mediation and moderation analyses, with bootstrap confidence intervals based on 5000 resamples. The results reveal that: (1) algorithmic anthropomorphism positively affects both algorithmic transparency and perceived algorithmic fairness; (2) algorithmic transparency has a significant positive effect on both perceived fairness and purchase intention; (3) perceived algorithmic fairness mediates the relationships between algorithmic anthropomorphism and purchase intention, as well as between algorithmic transparency and purchase intention; and (4) although the composite technology acceptance level (TAL) measure does not significantly moderate the anthropomorphism–purchase intention path (p = 0.075), disaggregating TAL into its sub-dimensions reveals that PEOU significantly moderates this relationship (p < 0.001), whereas PU does not (p = 0.199). The composite-TAL result is therefore not statistically supported, but the dimension-specific PEOU finding is robust. These findings offer theoretical contributions to AI-driven consumer behaviour research and practical implications for the design of algorithmic e-commerce systems in emerging digital markets.