-

This research provides a preliminary study on automating user persona generation via ChatGPT-4 with a case study on educational digital services. Data were collected from 40 undergraduate students at one of Thailand’s science and technology higher education institutes through questionnaires and semi-structured interviews, with participants grouped via K-means clustering based on similar responses. The study relied on ChatGPT-4 to generate personas based on empathy maps, jobs-to-be-done analysis, and Hofstede’s cultural dimensions. The validity of the generated personas was assessed using SHapley Additive exPlanations to identify key influencing features, revealing that integrating cultural contexts enhances persona specificity and relevance. The findings demonstrate that personas generated through this approach accurately reflect unique traits of target clusters, while significantly reducing the time and resources required. This study highlights a practical method for using ChatGPT-4 to generate personas, focusing on Thai science and tech undergraduates. It shows how tailored data collection improves ChatGPT’s understanding of specific user groups. Future research could expand sample sizes and explore broader contexts to enable generalization of the approach to become a framework that works across multiple target users and application contexts. Additionally, exploring methods like hierarchical K-means may improve clustering accuracy. Incorporating expert evaluations using the Persona Perception Scale could further ensure the real-world applicability.