Phishing attacks remain a growing global cybersecurity threat, with attackers constantly evolving their evasion techniques to bypass traditional detection methods. While machine learning shows significant promise in detecting phishing websites, it faces several implementation challenges. This study investigates the effectiveness of KBest feature selection in phishing website detection and compares various ensemble learning methods against traditional base learners. Our experimental results demonstrate that ensemble methods consistently outperformed their base learner counterparts without substantial computational overhead. Notably, the Bagging ensemble achieved superior performance with 98.66% accuracy on the dataset, highlighting the potential of ensemble approaches in enhancing phishing detection systems.