A Retrieval-Augmented Generation (RAG) Based Framework for Evaluating Urban Low-Carbon Governance and Its Implications for Sustainable Development

The transition to low-carbon urban environments is a cornerstone of global sustainability efforts. However, evaluating the diverse management efforts driving this transition is frequently hindered by fragmented and heterogeneous data. This study introduces a novel Retrieval-Augmented Generation (RAG) framework to systematically evaluate urban low-carbon governance. RAG represents a significant methodological innovation, integrating advanced information retrieval with generative artificial intelligence to provide a transparent and evidence-based assessment of policy implementation. The proposed framework is applied to evaluate the low-carbon energy practices of 296 Chinese cities. A critical finding is the systemic neglect of the “Check” phase across all urban tiers, posing a significant challenge to the long-term sustainability of carbon reduction initiatives. Conversely, high-performing cities are characterized by robust “Feedback” mechanisms. Policy priorities also vary significantly with city scale: larger cities emphasize strategic development and low-carbon transitions, while smaller cities focus on foundational planning and ecological preservation. This framework serves as a transparent, process-oriented tool for evidence-based low-carbon governance, thereby facilitating more resilient and sustainable urban pathways.