GenAI personalization: antecedents, outcomes, mediators, and moderators

Purpose As Generative Artificial Intelligence (GenAI) and Large Language Models (LLMs) increasingly shape personalization practices, scholarly attention to their emotional, relational and systemic implications has grown in the hospitality sector. To synthesize and integrate recent advances, this study constructs a conceptual framework that links not only antecedents and outcomes of GenAI personalization in hospitality activities, but also mediators and moderators of the relationships that drive these outcomes. Based on the gaps identified through this framework, this research develops directions for future research and theory development. Design/methodology/approach This research adopts a critical review approach, synthesizing literature published between 2020 and 2025 across human-centric AI, emotional computing and service research. A PRISMA-guided screening process and the Theory-Context-Characteristics-Methodology (TCCM) framework were used to identify dominant theories, methodological patterns and conceptual gaps. Insights from this synthesis informed the identification of aggregate themes and the development of an integrative conceptual framework, supported by causal loop modeling of key personalization dynamics. Findings GenAI capabilities, LLM-driven emotional adaptability, conversational naturalness and algorithmic transparency shape personalization fit and co-creation quality, which, in turn, activate both emotional engagement and personalization fatigue. These processes span five aggregate dimensions, namely, cultural, digital, economic, experiential and emotional-cognitive and are contingent on trust propensity, privacy concern and cultural orientation. While emotionally intelligent GenAI enhances perceived experience quality, satisfaction and brand loyalty, excessive or poorly calibrated personalization intensifies cognitive overload and fatigue. These portray personalization as a dynamic system in which reinforcing and balancing mechanisms jointly govern value creation, risk escalation and long-term relational outcomes. Research limitations/implications This research opens avenues for comparative, cross-cultural and longitudinal research designs that can deepen understanding of GenAI-driven personalization over time. It also highlights that GenAI personalization should be conceptualized not as a static technological feature but as an adaptive socio-technical system, requiring a shift beyond efficiency-centric perspectives toward frameworks that incorporate emotional resonance, cultural intelligence and relational sustainability in human–AI interactions. Originality/value This research advances GenAI personalization research by conceptualizing GenAI and LLMs as emotionally adaptive systems rather than purely functional technologies. It offers a novel conceptual framework that provides a foundation for future empirical testing and guides the responsible design of human-centered Gen AI personalization systems.