The rapid expansion of e-commerce and digital platforms has significantly transformed the way consumers gather information and make purchase decisions. In online environments characterized by information asymmetry and information overload, online revi...
The rapid expansion of e-commerce and digital platforms has significantly transformed the way consumers gather information and make purchase decisions. In online environments characterized by information asymmetry and information overload, online reviews have become one of the most influential sources of information for consumers. Traditionally, consumer-generated reviews have served as a form of electronic word-of-mouth (eWOM), helping potential buyers reduce uncertainty and evaluate product quality before purchase. However, recent advances in artificial intelligence (AI) technologies have introduced a new form of recommendation information: Al-generated reviews and recommendations. As consumers are increasingly exposed to both human-generated and AI-generated review content, understanding how different recommendation formats influence trust formation and purchase intention has become an important research issue.
Despite the growing prevalence of AI-generated recommendation systems, previous studies have primarily focused on the quantity, valence, and quality of reviews while paying limited attention to the source of recommendation information and the psychological mechanisms through which trust is formed. Furthermore, existing eWOM research has largely been grounded in human-generated communication contexts and has not fully explained how consumers evaluate and trust AI-generated information. To address these limitations, this study investigates how recommendation presentation format influences consumer purchase decisions and explores the role of recommendation trust as a key psychological mechanism.
This research integrates the Stimulus-Organism-Response (SOR) framework and the Technology Acceptance Model (TAM) to develop a comprehensive model of consumer decision-making in AI-enabled recommendation environments. Within the SOR framework, recommendation presentation format serves as the stimulus (S), recommendation trust represents the organism (0), and purchase intention functions as the behavioral response (R). In addition, TAM is incorporated to explain the cognitive evaluation process through perceived usefulness and perceived ease of use. Furthermore, self-control is introduced as a moderating variable to examine how individual psychological characteristics influence trust formation and behavioral responses toward different recommendation sources.
Based on this theoretical framework, the study compares two types of recommendation presentation formats: traditional human-generated reviews and AI-generated reviews. Human-generated reviews are characterized by experiential authenticity, emotional expression, and social credibility, whereas AI-generated reviews emphasize information efficiency, consistency, personalization, and algorithmic processing. This distinction provides an opportunity to examine whether consumers form trust through different psychological pathways depending on the source of recommendation information.
To empirically test the proposed framework, scenario-based experiments were conducted in online shopping contexts. Participants were randomly assigned to different review conditions and evaluated products based on either human-generated or AI-generated recommendation information. Multiple statistical analyses, including manipulation checks, t-tests, mediation analyses, logistic regression, and moderation analyses, were employed to examine the relationships among recommendation presentation format, trust, self-control, and purchase intention.
The findings reveal several important results. First, recommendation presentation format significantly affects consumers' trust perceptions. Human-generated reviews generally generate higher levels of trust because consumers perceive them as more authentic, emotionally relatable, and socially credible. In contrast, AI-generated reviews are evaluated based on their perceived usefulness, consistency, and technological reliability. These findings suggest that recommendation trust differs according to the source of information and that consumers rely on distinct trust formation mechanisms when evaluating human and AI-generated content.
Second, recommendation trust plays a significant mediating role in the relationship between recommendation presentation format and purchase intention. The results demonstrate that recommendation format influences consumers' trust perceptions, which subsequently affect their willingness to purchase. This finding supports the core assumption of the SOR framework by confirming that external information stimuli influence behavioral outcomes through internal psychological processes.
Third, cognitive evaluations derived from TAM contribute significantly to trust formation. Consumers who perceive recommendation information as useful and easy to understand are more likely to trust the recommendation source. This result highlights the importance of cognitive acceptance processes in technology-mediated consumption environments and suggests that trust formation in AI-based systems extends beyond traditional notions of social credibility.
Fourth, self-control significantly moderates the relationship between recommendation presentation format and recommendation trust. Consumers with relatively low levels of self-control tend to place greater trust in human-generated reviews, relying more heavily on social cues and experiential information. In contrast, consumers with higher levels of self-control exhibit relatively greater trust toward AI-generated recommendations, suggesting a stronger preference for systematic information processing and objective evaluation. These findings indicate that individual psychological characteristics influence how consumers interpret and trust recommendation information.
Theoretical contributions of this study are threefold. First, it extends traditional eWOM literature by incorporating AI-generated recommendation information and comparing trust formation mechanisms across human and AI sources. Second, it integrates SOR and TAM into a unified framework that explains how technological and psychological factors jointly influence consumer decision-making. Third, it identifies self-control as an important boundary condition that shapes consumer responses to recommendation information, thereby expanding current understanding of trust formation in digital environments.
From a managerial perspective, the findings provide practical implications for e-commerce platforms, marketers, and recommendation system designers. Platforms should recognize that consumers do not evaluate all recommendation information equally and that trust formation depends on both the source of information and individual consumer characteristics. AI-generated recommendation systems may become more effective when transparency, explainability, and perceived usefulness are enhanced. At the same time, maintaining elements of human authenticity and social credibility may help strengthen consumer trust and purchase intention.
In conclusion, this study contributes to the growing body of research on consumer behavior in AI-mediated environments by demonstrating how recommendation presentation format influences trust formation and purchase intention. As AI-generated content continues to expand across digital platforms, understanding the psychological mechanisms underlying consumer trust will become increasingly important for both theory and practice.