Prompts for planning-AI integration: LLM prompt design for supporting sustainable urban development

Large Language Models (LLMs), such as ChatGPT, are increasingly integrated into urban planning workflows to support tasks ranging from policy drafting to participatory engagement. Prompt engineering—the systematic design of instructions that guide LLM behaviour—has emerged as a critical factor determining the quality, relevance, and reliability of AI-generated outputs in planning applications. However, limited understanding of how prompts are constructed and adapted for planning contexts constrains the effectiveness, transparency, and reproducibility of these applications. This systematic review examines peer-reviewed studies to investigate prompt engineering applications in urban planning and adjacent domains. The study identifies seven standardised component categories and eight key prompting techniques, revealing distinctive typological patterns in prompt template structures. Based on these insights, the paper proposes a novel three-layer framework—task adaptation, component configuration, and enhancement—that supports the development of task-specific, modular prompts with high adaptability across diverse planning scenarios. This framework addresses current limitations static design and underdeveloped interaction mechanisms, enabling more context-aware and accountable LLM applications. In doing so, it supports the integration of AI into sustainable urban development by enabling more context-aware, accountable, and strategy-aligned applications of LLMs in planning workflows. By transforming ad-hoc prompting into structured methodology, this study provides foundations for reliable, transparent AI deployment in urban planning and establishes systematic design principles supporting sustainable urban development through effective human-AI collaboration.