Phishing URL detection has become a pivotal element of cybersecurity phishing attacks grow in complexity. This work introduces a mechanism designed to identify that uses the Gradient Boosting Machine Learning (ML) algorithm, achieving 98% accuracy. In our work, the algorithm was trained to distinguish between legitimate and harmful URLs using a large-scale Kaggle dataset that contained various features. The examined key metrics comprise URL length, domain age, subdomains, HTTPS presence, special characters, IP addresses, redirects, suspicious keywords, domain length, and Port presence, as well as the absence of www and the URL path length. These functionalities enable precise evaluation of URL security. The system’s user-friendly interface classifies URLs in real-time as either “safe” or “unsafe” with the percentage meter and also blocks unsafe URLs that go beyond a preset threshold. Furthermore, in order to achieve optimal performance, an intelligent caching mechanism block is used on previously analyzed URLs, which improves the response time. With respect to accomplishing precise predictions as well as handling complex datasets, the Gradient Boosting algorithm was chosen. Our proposed system combines feature engineering with lexical, domain, and network attributes for better detection. The Kaggle publicly available dataset was used for testing the model, and it proved to be effective and robust, which confirms the assessment. Overall, the proposed work shows the real implementation of ML in combating phishing effectively and provides a strong solution in automating phishing attempts due to high accuracy, easy access to the graphical interface, and optimized backend of the system.