The present study employed the draw-a-picture technique and epistemic network analysis (ENA) to reveal university students' viewpoints on ChatGPT-supported learning, as well as the conceptions, roles, and educational objectives of ChatGPT-supported learning among university students with different learning attitudes. The results showed that university students' conceptions of ChatGPT-supported learning were diverse and interrelated. Most students agreed that ChatGPT could facilitate learning anytime and anywhere, and held positive emotions and attitudes. In the conceptions of ChatGPT-supported learning, there were significant differences in the four categories of locations, learning content, learning activities, and objects between the low learning attitude (LLA) and the high learning attitude (HLA) groups. Besides, the HLA group preferred to use ChatGPT for specific learning content, and showed higher levels of engagement in discussions and consultations. The LLA group was inclined to use ChatGPT at home or in places with traditional classroom equipment for completing reports or assignments. Most students in the HLA group regarded ChatGPT as a tutor, while those in the LLA group viewed it as a tool. Based on the above findings, this study puts forward some suggestions for how to implement ChatGPT-supported learning in higher education as a reference for educators, researchers and training developers.