Phishing websites pose a growing threat to internet users, and traditional detection methods like blacklisting or relying on SSL certificates fail to keep pace with rapidly evolving cyberattacks. This study introduces a new approach that leverages XGBoost, a powerful machine learning algorithm, combined with the firefly algorithm for hyperparameter optimization in phishing detection. Inspired by the flashing behavior of fireflies, the firefly algorithm fine-tunes critical hyperparameters like learning rate and maximum tree depth, enhancing XGBoost’s accuracy and ability to learn patterns without overfitting. This method balances exploring new solutions with refining the best ones, improving classification performance. This integrated approach provides an efficient and reliable solution for detecting phishing websites, strengthening cybersecurity in the fight against online threats.