<p>Social networks have gained widespread attention as an alternative data source for credit assessment. This study constructs a social network credit dataset using real microcredit data from rural China, examining the role of network information in rural residents’ digital credit assessment from both risk disclosure and risk prediction perspectives. The findings show that social network information complements traditional credit information, and residents with higher centrality tend to have lower default probabilities. Integrating social network features into machine learning-based credit assessment models enhances the accuracy of rural residents’ credit evaluations, particularly improving the identification of non-defaulters among thin-file and subprime borrowers. These findings provide valuable insights into the combination of traditional credit information with alternative data sources, enabling more accurate and inclusive credit assessments in rural areas.</p>