رفتار خرید مصرف‌کنندگان در مواجهه با اینفلوئنسرهای مجازی مبتنی بر هوش مصنوعی: نقش تعامل پاراسوشال، اعتبار منبع و ادراک انسانی

Abstract

The rapid expansion of digital environments has led to the growing presence and influence of AI-based virtual influencers (VIs) across social media platforms. These computer-generated characters, designed with humanlike physical and behavioral features, increasingly shape consumer engagement and purchasing decisions. This study examines how perceived anthropomorphism influences parasocial interaction, source credibility, consumer attitudes, purchase intention, and actual purchasing behavior in the context of AI-powered virtual influencers. Using a descriptive–survey methodology and structural equation modeling (SEM), data were collected from followers of virtual influencers through an online questionnaire (n=386). Results indicate that perceived anthropomorphism significantly enhances parasocial interaction and source credibility; however, its direct effect on attitudes was not significant. Furthermore, parasocial interaction positively affects attitudes and source credibility but does not significantly influence purchase intention. Source credibility emerged as the dominant predictor of purchase intention, which subsequently drives actual purchase behavior. These findings confirm that the only significant pathway from anthropomorphism to purchase-related outcomes is mediated through source credibility. The study contributes to the emerging literature on virtual influencer marketing by offering an integrated model that combines psychological and technological factors influencing consumer decision-making.

Introduction

The evolution of influencer marketing has transformed digital advertising strategies, especially as AI-generated virtual influencers gain traction among brands and consumers. With high engagement potential and immunity to risks faced by human influencers (e.g., scandals, inconsistency), VIs have become a preferred option for major global brands. Despite their rising popularity, research on consumer psychological responses to VIs remains limited. Previous studies highlight mixed reactions: some consumers show increased engagement, whereas others report discomfort or distrust rooted in the “uncanny valley” effect. Given this divergence, understanding psychological mechanisms—such as parasocial interaction, perceived human likeness, credibility, and attitudes—has become essential. Moreover, few studies have integrated these variables into a comprehensive model that explains how VIs influence consumer purchase behavior. This study fills this gap by examining anthropomorphism as an antecedent to key psychological constructs and behavioral intentions, grounded in established theories such as Source Credibility Theory and the Theory of Planned Behavior.

Materials and Methods

This research is applied in purpose and descriptive–survey in methodology. The statistical population consisted of followers of AI-based virtual influencers. Since the population size was unknown, Cochran’s formula was used to determine the minimum sample size (n=386). Ultimately, 386 valid responses were collected through online distribution. Non-probability convenience sampling was applied.

The measurement scales were adapted from validated instruments: perceived anthropomorphism and parasocial interaction (Um, 2023), source credibility (Chetioui et al., 2019), attitude, subjective norms, perceived behavioral control, and purchase intention (Zaharia & Asici, 2024), and actual purchase behavior (Venciute et al., 2023). All items were measured using a 6-point Likert scale. Reliability was assessed using Cronbach’s alpha and composite reliability (CR), both exceeding 0.70. Convergent and discriminant validity were confirmed through AVE and Fornell–Larcker criteria. Data analysis employed structural equation modeling using SmartPLS 4. Model fit was evaluated through R², Q², and SRMR indices.



Discussion and Results

The structural model supported 9 out of 11 hypotheses. Perceived anthropomorphism significantly influenced parasocial interaction and source credibility but did not significantly affect attitude—suggesting that excessive human likeness may trigger uncanny valley responses, diminishing affective evaluation. Parasocial interaction significantly enhanced both source credibility and attitude but showed no direct impact on purchase intention. This implies that emotional bonding with a virtual influencer alone is insufficient to stimulate purchase intention unless accompanied by perceived expertise, trust, or attractiveness.

Source credibility demonstrated a strong positive effect on purchase intention, confirming its central role as a mediator. Additionally, both perceived behavioral control and subjective norms significantly shaped consumer attitudes, supporting the Theory of Planned Behavior. Attitude strongly predicted purchase intention, which in turn had a powerful and significant effect on actual purchase behavior.

Overall, the results reveal that source credibility is the strongest determinant of purchase-related outcomes, and the only meaningful path from anthropomorphism to purchase behavior is through credibility, not parasocial interaction or attitude. These findings align with theories emphasizing trust and perceived expertise as critical for persuasion in digital environments.

Conclusions

This study enriches the understanding of how AI-generated virtual influencers affect consumer psychology and behavior. Findings highlight that anthropomorphic design enhances psychological engagement but may also trigger negative reactions if overly realistic. Importantly, credibility—not emotional bonding—is the key mechanism translating anthropomorphism into actual consumer behavior.

The results offer practical implications for brands and developers: VIs should balance realism and distinctiveness to avoid uncanny valley discomfort; credibility-enhancing attributes such as expertise, consistency, and trustworthy narratives should be emphasized; and designers should consider cultural and social norms that influence consumers’ perceptions of AI-based influencers.

Future research should explore cross-cultural differences, develop multi-dimensional frameworks of anthropomorphism, and test virtual influencers across diverse digital ecosystems such as the metaverse.