Phishing attacks pose a persistent and growing threat to cybersecurity, targeting users across various communication platforms. Traditional detection systems often struggle to keep up with the evolving sophistication of phishing techniques. This research presents an AI-driven phishing detection system that provides real-time protection across diverse channels, including email and popular messaging applications like WhatsApp and Facebook Messenger. Our approach combines DistilBERT for efficient, text-based phishing detection with XGBoost for URL analysis, creating a robust, multi-layered defense. Leveraging state-of-the-art natural language processing and machine learning algorithms, the hybrid model demonstrates high detection performance, achieving accuracy, precision, and recall rates up to 89% with balanced false positive and negative rates, as shown by experiments with different weight ratios (e.g., 4:1 for BERT and 7:4 for DistilBERT). The system’s scalability and adaptability make it a viable, real-time solution to counter sophisticated phishing tactics across multiple platforms, representing a significant advancement in the security of digital communications.