This paper argues that while generative AI, large-scale language models, and agent AI have evolved independently in urban planning, their convergence presents exciting opportunities for AI urban planners. Existing research conceptualizes urban planning as a generative AI task, suggesting ways for AI to synthesize land-use configurations and reconstruct automated urban designs within geospatial, social, and human-centered constraints. However, this research highlights the need for strong pre-defined generative structures (e.g., adversarial generator-discriminator, forward and backward diffusion structures, and hierarchical zone-point-of-interest generation structures) and overlooks the use of tools developed by urban planning experts. To address these limitations, this paper proposes a future research direction, Agent City AI Planner, calling for a novel integration of agent AI and participatory urbanism.
