Cybersecurity faces persistent threats from phishing attacks, prompting the need for robust URL detection systems. This research explores the efficacy and interpretability of Random Forest and Artificial Neural Network (ANN) models, employing SHAP (SHapley Additive exPlanations) for feature importance analysis. Using these models, phishing URLs were classified, and SHAP facilitated the understanding of feature significance in model decision-making. Comparative analysis revealed distinct feature preferences between models. Random Forest emphasized Google index, page rank, and web traffic, while ANN prioritized page rank, Google index, and URL structure attributes. These findings underscore the models' divergent feature inclinations, providing actionable insights for feature selection and model enhancement in phishing URL detection.