Recruitment scams are a growing problem, and their consequences involve various financial and emotional aspects for applicants. This systematic review evaluates the efficiency of using Machine Learning (ML) and Deep Learning (DL) techniques to detect recruitment scams in online job advertisements. This study analyzed current approaches, identified key challenges, and explored future directions. The analysis proposed for this review begins with a methodology in which an exhaustive search of the Scopus database was performed using a structured search equation according to the PICOC methodology. The inclusion and exclusion criteria were applied to select relevant studies. Results were then presented showing that models based on Random Forest, XGBoost, and Bi-LSTM demonstrated the best performance, with accuracy rates between 91.8% and 99.9%. Data imbalance management techniques, such as SMOTE and ADASYN, significantly improved performance. In addition, persistent challenges such as model generalization, rapidly evolving fraud tactics, and the need for more representative data sets, were identified. Finally, the conclusions section synthesizes the results, where we identified that the contribution of ML and D L proved to be effective tools for the detection of online recruitment scams, with significant advances in accuracy and handling of complex data.