Pan evaporation is a key indicator in hydrological modeling and water resources management. Therefore, developing reliable models for its estimation is of great importance. Conventional empirical and physical methods often face limitations due to their extensive data requirements and sensitivity to climatic conditions. In response to these limitations, recent research has turned to machine learning and deep learning algorithms. These approaches have emerged as efficient tools for modeling and predicting pan evaporation. The objectives of this study are to analyze scientific trends and publication patterns using bibliometric analysis, and to identify and compare pan evaporation modeling approaches through a structured synthesis of the literature. Following an initial screening of 205 articles, 120 relevant papers were selected and analyzed. In the bibliometric section, the temporal trend of publications, geographical distribution, collaboration networks, and keyword analysis were examined. The results indicate a significant increase in the volume of publications and scientific collaborations since 2019, particularly in India, Iran, and China. Keyword analysis reveals that studies have focused on concepts such as "pan evaporation," "machine learning," and "artificial neural network." In the structural synthesis of the research, ANN, ANFIS, and SVM models were identified as the most widely used approaches, while hybrid models based on metaheuristic algorithms were introduced as a recent growing trend. Despite these advances, challenges remain, including limitations in station-based data, weak international collaborations, and insufficient attention to model uncertainty assessment. By providing a comprehensive overview of the scientific landscape and existing knowledge gaps, the findings of this study can help chart a course for future research aimed at improving the accuracy and robustness of pan evaporation models.