Phishing remains a growing cybersecurity threat, exploiting human vulnerabilities through deceptive communication tactics across platforms like email and social media. Traditional defenses such as firewalls and antivirus software are insufficient against these social engineering attacks, which now use AI-generated content and sophisticated deception to bypass detection. This research presents a lightweight browser-based prototype for real-time phishing detection, using machine learning techniques based on TF-IDF vectorization and a Naive Bayes classifier. Unlike large language model-based tools, the proposed solution is optimized for efficiency, speed, and accessibility. With a strong testing accuracy of 91% on independent datasets, the tool provides explainable results, supports email and attachment analysis, and offers educational feedback to improve user awareness. It is designed to run in real-time browser environments, balancing usability, transparency, and performance.