<p>ChatGPT has quickly attracted significant attention in the education domain. However, the education communities still lack a clear understanding of ChatGPT usage and the acceptance of this technology requires further investigation. Therefore, this study investigates factors affecting academic staff’s willingness to use ChatGPT for teaching and learning. Compared to the available studies that have not considered the unique attributes of ChatGPT in developing adoption models, this study used AI device use acceptance (AIDUA) as the foundation model. An online survey was conducted to collect data from academic staff with experience using ChatGPT for teaching and learning in public Malaysian universities. A total of 418 valid responses were obtained. This research then used the structural equation modeling-artificial neural network (SEM-ANN) approach to test the hypotheses and reveal the level of their importance toward willingness to use ChatGPT. SEM results showed that social influence and hedonic motivation positively impacted performance expectancy. Effort expectancy was positively influenced by anthropomorphism and hedonic motivation while negatively impacted by social influence. Both effort and performance expectations contributed to enhancing emotions. Willingness to use was significantly predicted by emotions and anthropomorphism. The results from the ANN underscored the importance of hedonic motivation as the most significant factor influencing both performance and effort expectancy. Additionally, anthropomorphism was found as the primary predictor, followed by emotions, in determining the willingness to use ChatGPT. These findings carry actionable implications for technology developers, academic institutions, and educators by offering valuable insights to support the adoption of large language models (LLMs), including ChatGPT, among academic staff.</p>