In an era characterized by the deep integration of digitalization and intelligent technologies, generative artificial intelligence (GAI) is reshaping the ecology of higher education in unprecedented ways. Owing to its inherent complexity, practice-oriented nature, and interdisciplinary characteristics, undergraduate architectural education can no longer be fully supported by traditional pedagogical models in response to emerging demands such as sustainable design, digital twins, and intelligent construction. Based on cross-sectional survey data from 1121 architecture undergraduates across eight universities in Wuhan, Hubei Province, this study proposes the Generative AI-enabled Learning Reshaping Association Model (GAI-LRM) and employs partial least squares structural equation modeling (PLS-SEM) to examine the statistical relationships between the variables. The generative AI tools investigated include text-generation tools such as Kimi AI, Doubao and Seedance, as well as image and design generation tools like Midjourney, Stable Diffusion, Forma AI, and ArkoAI. The results indicate that system-generated content quality, system quality, and task–technology fit are all significantly and positively associated with learning reshaping. Learning relationship reshaping and cognitive flexibility demonstrate positive indirect associations within the relevant pathways, whereas technology dependence shows a negative indirect association. Furthermore, there is a significant association between students’ foundational knowledge in the subject and certain variables. These findings reveal the multifaceted connections between the application characteristics of generative AI and changes in the learning processes of architecture undergraduates; they provide empirical insights for optimizing human–AI collaborative learning, critical design reviews, and tiered instruction in design studios at universities in Wuhan, while also establishing a theoretical framework for future cross-regional, longitudinal, and experimental studies. We situate these findings within a core framework of contemporary architectural scholarship, where mainstream architectural education continues to privilege image-driven representation and adherence to established stylistic paradigms, even as a parallel scientific research movement harnesses artificial intelligence to reshape fundamental design principles. Viewed from this perspective, our results reveal not only the current state of technology adoption but also the underlying mechanisms at play. Specifically, technological reliance diminishes cognitive flexibility, while deep disciplinary literacy constitutes the critical differentiator between uncritical replication and deliberate application. Consequently, we argue that architectural education should not merely incorporate GAI within existing visual paradigms but should instead steer it toward science-based, human-centric design principles.
Peng et al. (Tue,) studied this question.