Artificial intelligence for sustainable architectural design

Amid global challenges of resource depletion, climate risk, and urban health inequality, Artificial Intelligence for Sustainable Architectural Design (AI4SAD) has become a key driver of the shift from experience-based to intelligence-driven design. Drawing on a systematic review of 408 studies, this work is the first to map the global spatiotemporal profile of AI4SAD across design stages, sustainability goals, and algorithmic applications. The findings indicate a significant transparency deficit: only a few studies disclose data sources, model parameters, or performance gains. This gap constrains reproducibility, cross-scenario transfer, and cumulative knowledge. We propose a three-stage development framework and identify an ongoing transition from level 2 (specialized models) to level 3 (foundation models). Furthermore, a roadmap is established to guide future advancements in generalizability, autonomy, and interpretability, promoting the responsible application of AI in sustainable architectural design. Broader context: Urban sustainable transition shapes energy, climate, and public health. AI transforms sustainable architecture design by boosting efficiency and optimizing resources for UN goals. Yet data and transparency gaps limit its potential. We advocate for transparent, generalizable methods, regional support, and explainable, responsible AI.