A study of employees' perceptions of ChatGPT as an emotional support tool

As employee well-being receives increasing attention and the use of AI chatbots such as ChatGPT becomes more widespread, their potential role as emotional support tools for mitigating burnout remains underexplored. This study examines employees’ perceptions of ChatGPT as a source of emotional support and investigates how it may be used to alleviate burnout. Adopting a qualitative research design, the study employs thematic analysis of semi-structured interviews with eight employees from private technology firms in Bangkok, Thailand.The findings indicate that adoption of ChatGPT for emotional support is positively influenced by facilitating conditions, social influence, trust, and perceived anthropomorphism, while perceived risk exerts a negative effect. The analysis further reveals that critical evaluation mediates the relationship between performance expectancy and trust. In addition, perceived interaction complexity negatively moderates the relationship between effort expectancy and adoption. Two additional themes—24/7 accessibility and perceived emotional connection—emerged as salient drivers of adoption, as participants viewed ChatGPT as a readily available source of consultation, comfort, and encouragement.This study contributes to the Unified Theory of Acceptance and Use of Technology (UTAUT) by demonstrating that its core constructs—performance expectancy, effort expectancy, facilitating conditions, and social influence—are insufficient to fully explain the adoption of AI-based emotional support tools. Factors such as perceived emotional connection, continuous accessibility, trust, perceived risk, and perceived anthropomorphism play critical complementary roles. The study also offers practical implications for employees, organizations, and AI developers seeking to harness AI’s benefits for burnout mitigation and workplace well-being.While limited by a small, cross-organizational sample, this research lays the groundwork for future studies employing larger, single-organization samples, experimental or longitudinal designs, and quantitative or mixed-methods approaches, as well as more systematic examination of ethical and psychological concerns surrounding AI-mediated emotional support.