Phishing attacks remain a critical cybersecurity challenge, targeting individuals and organizations by impersonating legitimate entities to steal sensitive information. Existing detection mechanisms, such as blacklist-based filtering and rule-based systems, are limited in their ability to identify newly generated or obfuscated phishing URLs. This paper proposes PhishVision, a dual-layer phishing detection framework that integrates machine learning-based URL analysis with computer vision-driven visual verification. The system is deployed as a real-time browser extension with a FastAPI backend for efficient processing. The first layer utilizes a Random Forest classifier trained on engineered URL features to predict phishing likelihood. The second layer employs Optical Character Recognition (OCR) to extract textual content from webpage screenshots and detect inconsistencies between claimed brand identities and actual domain names. A decision engine combines outputs from both layers to produce a final classification with confidence scores. Experimental results indicate that the proposed approach improves detection accuracy and robustness against visually deceptive phishing attacks, making it suitable for real-time applications.