Abstract In recent years, microplastic contamination in freshwater lakes has become a significant environmental concern. Despite this, there remains a lack of comprehensive understanding of the distribution patterns and regional characteristics of microplastic loads in global lacustrine environments under a unified standard. To address this gap, our study utilizes Machine Learning (the random forest algorithm), combined with number-to-mass transformation techniques to generate a global prediction. The results indicate an average microplastic concentration of 0.57 items/m3 in lakes and reservoirs worldwide, with an accumulated microplastic load of 10167 tons within top 20 m of water—equivalent to 508 million plastic bottles. The primary sources of microplastics are linked to agricultural land use and the proportion of urban areas within watersheds. Notably, the highest microplastic loads are observed in North America, Africa, and Asia, though the contributing factors vary, including concentration-dependent and area-dependent influences, as well as differences in shape composition. These findings provide valuable insights that can guide the development of targeted policies to effectively mitigate microplastic pollution in freshwater ecosystems.