Right and Wrong: Harnessing visual geodata mining methods to find patterns in LLM generated geodata

The growing interest in leveraging Large Language Models (LLMs) as sources of spatial data necessitates the development of scalable, comprehensive, and standardized methods for assessing their quality. This study focuses on evaluating the relevance of spatial data generated by an LLM. The analysis combines visual inspection with a distance-based metric to assess the spatial relevance and accuracy of the generated data, providing a scalable approach for systematic quality evaluation.