Generative AI for Urban Co-Design: A Survey and the SPICE Conceptual Framework for Participatory 3D Scene Editing

Urban co-design is increasingly shifting toward participatory processes involving citizens, planners, and diverse stakeholders. Recent advances in generative artificial intelligence (GenAI) create new opportunities to support these processes through intuitive visual and spatial content generation. This paper surveys GenAI methods and tools for urban co-design, focusing on participatory 3D scene generation and editing. We review user-facing platforms, structured generative workflows, and GenAI-based 3D generation/editing approaches, highlighting their limitations for non-expert participation. Drawing also on evidence from the Indre participatory design project, we identify a key gap: current systems remain fragmented across analysis, generation, visualization, and editing stages, and lack integrated workflows for iterative and transparent 3D co-design. To address this, we outline the SPICE conceptual framework, combining multimodal interaction, human-in-the-loop refinement, NBS-driven prompt grounding, and before-after evaluation. The proposed direction aims to support more inclusive, transparent, and sustainability-driven urban co-design.