Grasslands, a vital ecosystem and component of the global carbon cycle, play a significant role in evaluating ecosystem health and monitoring the global carbon balance. In this study, based on the Carnegie–Ames–Stanford Approach (CASA) model, we estimated the Net Primary Productivity (NPP) of grasslands in northern Shaanxi from 2000 to 2020. Employing trend analysis, stability analysis, multiple regression analysis, and residual analysis, the research examined the dynamic changes of grassland NPP and its response to climatic and human factors. Key findings include: (1) Grassland NPP showed a significant increasing trend during 2000–2020, with high-coverage grasslands showing a higher rate of increase than medium and low-coverage grasslands. (2) Most grasslands (>90%) exhibited unstable growth and high NPP fluctuation. (3) While temperature, precipitation, and radiation undulate, the trends were not significant. Rainfall and radiation emerged as dominant factors affecting NPP, with temperature suppressing NPP increase to some extent. (4) Policies like returning farmland to grassland had a positive impact on grassland recovery, vegetation productivity, and regional ecosystem health.