Generative AI for Urban Planning and Design: Progress Review and Future Perspectives

The integration of generative artificial intelligence (GenAI) into urban planning and design has rapidly advanced as a key research frontier in recent years. This study reviews the application and emerging trends of GenAI in different stages of planning and design, including theoretical understanding, spatial analysis, and generation and evaluation of planning and design. Specifically, ① theoretical understanding: GenAI can construct multimodal knowledge graphs that support a more systematic understanding of fragmented knowledge in planning and design by integrating heterogeneous textual, visual, and spatial data. ② Urban spatial analysis: GenAI can enhance analytic capacity and inclusiveness in spatial analysis. It enables efficient interpretation of current socioeconomic and spatial conditions from multimodal data and can simulate the reasoning of multiple stakeholders, including experts and the public. Although limited in mechanistic, rule-based analyses, it can be extended via prompt engineering and tool use, lowering technical barriers. ③ Planning and design generation: GenAI can assist practitioners in drafting text, generating design images, and producing simple three-dimensional models. However, it is not yet capable of independently producing comprehensive, regulation-compliant planning documents or spatial layouts. Thus, it should be regarded as a supplementary tool rather than a replacement for human expertise. ④ Planning and design evaluation: Through domain adaptation and knowledge integration, GenAI can evaluate planning texts, spatial performance, ecological performance, and other multicriteria dimensions. It also supports multistakeholder assessments via large language model-based agentic workflows, but does not yet reliably automate the iterative optimization of proposals. Although the use of GenAI in this field is still in its early stages, it demonstrates cross-process potential across the entire planning and design workflow. It is expected to accelerate the shift toward computational urban science, move practice from experience-oriented to engineering-oriented approaches, and enhance the efficiency and quality of public participation, thereby better aligning planning outcomes with diverse societal needs.