Generative AI stickiness: A Conceptual Model Integrating AI Characteristics, Cognitive Dissonance, and Motivational Differences

User stickiness reflects users’ sustained engagement with and reliance on a system, and is therefore central to the long-term viability of generative artificial intelligence (GenAI). Drawing on Cognitive Dissonance Theory, this study develops a conceptual understanding of how two core AI characteristics—perceived intelligence and perceived anthropomorphism—shape user stickiness through the regulation of cognitive dissonance, while considering motivational differences—specifically utilitarian and hedonic motivations—as important boundary conditions. In early interactions, GenAI may elicit cognitive dissonance when output deviations such as hallucinations or logical inconsistencies violate users’ performance expectations. As interactions accumulate, improvements associated with perceived intelligence, including enhanced response quality and problem-solving capability, may facilitate cognitive reconciliation and alleviate psychological discomfort. Similarly, anthropomorphic cues may initially induce dissonance by fostering unrealistically high expectations, yet with repeated interaction, perceived anthropomorphism can enhance users’ tolerance for AI imperfections and emotional adaptability by enabling human-like interaction and emotional connection. These regulatory processes may vary across users with different motivational orientations. Based on these arguments, this study proposes a conceptual model of “AI perceived characteristics → cognitive dissonance → user stickiness”, in which motivational differences shape users’ psychological responses to dissonance. By elucidating the dynamic and dual roles of perceived intelligence and anthropomorphism, the proposed framework extends cognitive dissonance theory to continuous human–AI interaction contexts and offers conceptual guidance for the design of interactive intelligent systems aimed at sustaining user stickiness.