Extending UTAUT2 to explore intention to use ChatGPT for travel planning: a hybrid PLS-ANN approach

Purpose Generative artificial intelligence (AI) can revolutionize the tourism and hospitality industry by offering personalized recommendations and simplifying the process of obtaining travel information. This study investigates the critical drivers of ChatGPT usage for travel planning. The study extended the “extended unified theory of acceptance and use of technology” UTAUT2 by assessing the direct and moderating impacts of personal innovativeness and risk aversion. Design/methodology/approach A sample of 410 respondents, consisting of Malaysians aged 18 and above, took part in an online survey. Before answering the survey, the respondents were given the opportunity to practice obtaining travel information using ChatGPT. The collected data was then analyzed using a hybrid approach that combined “partial least squares” (PLS) and “artificial neural network” (ANN) techniques. This analysis aimed to test the significance of direct and moderating effects as well as rank the influential drivers. Findings PLS results revealed that performance expectancy, hedonic motivation, facilitating conditions, personal innovativeness and risk aversion significantly influence the intention to use ChatGPT for travel planning. Personal innovativeness moderates negatively the impact of effort expectancy on ChatGPT usage. Risk aversion moderates negatively the effects of social influence, effort expectancy and hedonic motivation. ANN results imply that performance expectancy is the most influential driver of ChatGPT usage, followed by hedonic motivation and personal innovativeness. Practical implications The findings provide insights for generative AI developers and tourism and hospitality service providers on how to trigger the use of ChatGPT for travel planning. Originality/value The study contributes to the literature by (1) assessing the drivers of intention to use ChatGPT for travel planning, (2) extending the UTAUT2 model and (3) using a hybrid PLS-ANN approach.