Monitoring the use of geography textbooks in secondary schools is a key component of the textbook governance system. To address the multiple challenges confronting current monitoring practices—including conflicts in functional positioning, technical complexity, and institutional bottlenecks—this study, grounded in an indicator system for monitoring the use of secondary school geography textbooks, develops a large-language-model-based agent for textbook-use monitoring and proposes an overall architecture consisting of materials upload–intelligent analysis–feedback optimization. An application case based on a lesson design on Karst Landforms shows that the agent can eff ectively identify evidence of textbook use while demonstrating strong explainability and clear awareness of evidential boundaries. In addition, through layered feedback and follow-up monitoring, the system preliminarily enables a shift in textbook-use monitoring from result presentation to improvement-oriented support.