Background: Urolithiasis is one of the most common urological diseases worldwide. One of the most common therapy, extracorporeal shock wave lithotripsy (ESWL), has a high failure rate. The failure rate can be significantly reduced by identifying the candidates most likely to benefit from ESWL, for example, by using machine learning (ML) algorithms. Decision tree analysis (DTA), artificial neural networks (ANN), and random forests (RF) represent a few of the machine learning approaches employed to forecast the stone-free outcome following ESWL. Methods: 219 studies were searched through six electronic databases (CENTRAL, MEDLINE, EMBASE, EBSCO, Proquest, SCOPUS). We employed the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines and adhered to the Standards for Reporting Diagnostic Accuracy Studies (STARD). To evaluate the potential bias in all the studies, we utilized the Quality Assessment of Diagnostic Accuracy Studies (QUADAS-2) tool. Results: 41,484 patients from 11 studies were included. The ML models highlight varying levels of diagnostic precision, with sensitivity spanning from 35-96%, and specificity ranging from 63-98.4%, and area under the curve falling between 0.49-0.96. It is shown in this study that the accuracy of RF and DTA in predicting stone-free status is superior than ANN. Conclusion: ML is a comparable predictive method to statistical analysis in predicting stone-free status. Random forest method and DTA are superior MLs compared to ANN. Stone size, density, and 3D texture analysis are the most important variables to be considered in the ML models and should be included in the models to ensure accuracy of stone-free status prediction. Copyright: © 2025. F et al.