Generative AI and the Transformation of Travel Planning: A Serial Mediation Model of Tone Appropriateness, Trust, and Time Efficiency in ChatGPT-Assisted Travel Decision-Making

This study examines how travellers’ use of ChatGPT—a generative artificial intelligence system—relates to travel decision-making through perceived tone appropriateness, trust in non-human assistance, and perceived time efficiency. Drawing on the Stimulus–Organism–Response (S-O-R) framework, the study conceptualises ChatGPT use as the focal technological stimulus, the three evaluations as distinct organismic states, and travel decision-making as the response. Data were collected through an online survey of 784 respondents who had used ChatGPT for travel-related planning. The proposed model was assessed using partial least squares structural equation modelling (PLS-SEM) with bootstrapping. The findings indicate that ChatGPT use is positively associated with travel decision-making as well as with perceived tone appropriateness, trust, and perceived time efficiency. Each of the three evaluations is also positively related to travel decision-making. In addition, tone appropriateness is positively associated with trust, while trust is positively associated with perceived time efficiency. The specific indirect associations through each mediator are significant, while the theoretically specified serial indirect association is statistically significant but small (β = 0.019). Comparisons of parallel, reduced sequential, and full serial specifications, alongside a linear-model predictive benchmark, indicate that the additional explanatory and predictive gains for travel decision-making are marginal, with greater improvements for the intervening constructs. These findings highlight distinct but statistically connected evaluations rather than a dominant serial mechanism.