Individual differences–specifically task engagement (TE)–in task-based language teaching have recently stimulated scholarly inquiry among second language researchers and educators. Nevertheless, most studies focus on learners’ engagement in human-designed tasks; few studies have employed ChatGPT as a tool to develop task-based materials to facilitate language learning. This study aimed to narrow the gap by exploring how Vietnamese English majors engaged with tasks generated by ChatGPT in accordance with a text-driven task-based approach (Tomlinson, Brian (ed), Developing materials for language teaching, Bloomsbury Publishing, 2023). Two 50-min speaking task-based lessons generated by ChatGPT based on preselected texts were delivered to thirty English sophomores at a public university. An explanatory sequential design was used to gather data, including a questionnaire on task engagement after each lesson, and a follow-up semi-structured interview to delve into dimensions of engagement. Findings showed that participants were highly engaged in the speaking tasks, consistently in four dimensions across the two lessons. Behaviourally, the learners tried to complete and stay focused on their tasks despite challenges; cognitively, they attempted to connect relevant information for task completion, while emotional engagement was reflected in their interest in trendy topics while participating in role-play and opinion exchanges. Social engagement involved seeking help with new words and ideas for speaking. The study implies that teachers can use AI for supporting task design and implementation to promote learner engagement.