Mapping News Geography: A Computational Framework for Classifying Local Media Through Geographic Coverage Patterns

This study introduces a novel computational framework for defining local news outlets through their geographic coverage patterns. This approach addresses the growing disconnect between legacy spatial markers (e.g., newsroom location or circulation cut-offs) and the geographic dimension of news coverage amidst media ownership consolidation and digital transformation. We develop a four-step pipeline consisting of data sampling, geoparsing, feature engineering, and clustering analysis. Our approach employs large language models for toponym disambiguation and develops eight spatial metrics across four dimensions: spatial extent, administrative reach, spatial heterogeneity, and distance decay. We test this pipeline on a sample of more than 465,000 articles from 360 UK digital local news outlets. Clustering analysis of more than 1.3 million locations reveals six distinct outlet types, ranging from hyperlocal and metropolitan to national outlets. This classification reflects different scales and structures of news provision in the UK’s evolving media landscape. The method offers a scalable, open-source approach for mapping local news coverage and understanding the scope of local news providers as a function of their coverage, with implications for media geography and policy, ownership studies, and the computational humanities.