This paper focuses on leveraging machine learning models to detect phishing websites by analyzing both URL-based and content-based features. Phishing presents a significant cybersecurity threat, as attackers deceive users by mimicking legitimate sites to steal sensitive information. The study assesses several machine learning models, including Decision Trees, Random Forests, Support Vector Machines (SVM), Gradient Boosting, XGBoost, K-Nearest Neighbors (KNN), and Naive Bayes, across four datasets. Comprehensive preprocessing was conducted to enhance model performance, incorporating techniques like handling missing data, removing duplicates, dimensionality reduction, outlier detection, feature scaling, and feature engineering. To evaluate the models’ effectiveness, key metrics such as accuracy, precision, recall, F1-score, and ROC AUC were employed. Results indicate that ensemble methods like Random Forest achieved an impressive accuracy of 96.88%, while XGBoost reached 95.90%. SVM, and Gradient Boosting demonstrated balanced performance with accuracies around 95%. Conversely, simpler models like KNN and Naive Bayes struggled with overfitting, particularly on smaller datasets, where accuracy dropped below 83%. Overall, this study emphasizes the efficacy of advanced models in phishing detection and highlights the essential role of preprocessing in enhancing model robustness and accuracy.