This study investigates the impact of the family doctor contracted service system on the health of migrants in China, utilizing data from the 2018 China Migrants Dynamic Survey (CMDS). The study employs a double machine learning model to estimate the effect of family doctor contracted services (FDCS) on migrants' self-rated health (MSRH). The sample consists of 137,851 migrants, with family doctor service contract status, health education, and health records as key variables. To address potential endogeneity issues, an instrumental variable approach using the regional family doctor contracting rate was implemented. Mediation analysis was conducted to examine the roles of health education and health records in this relationship. The findings indicate that FDCS significantly improve MSRH. This positive effect is robust across various machine learning models, including Lassocv, Random Forest, and Gradient Boost. The instrumental variable approach confirms the validity of these results, mitigating concerns about endogeneity. Mediation analysis reveals that the positive impact of FDCS on MSRH is fully mediated by health education and health records, highlighting their critical roles in enhancing health outcomes. The FDCS play a crucial role in improving the health of migrants by providing continuous and comprehensive care. Enhanced health education and effective health records management are significant pathways through which these services exert their positive effects. Policy recommendations include expanding access to family doctor services, enhancing health education programs, and improving health records management to optimize healthcare delivery for migrants. Future research should consider longitudinal studies to further validate these findings and explore their applicability to specific subgroups or regions.