This paper presents a hybrid framework for automated phishing detection, which combines multiple models to enhance both effectiveness and robustness. Phishing attacks are a critical cybersecurity threat, and current systems often rely on single-analysis models, making them vulnerable to sophisticated bypass attempts. The proposed hybrid framework addresses these limitations by integrating different models, allowing for more comprehensive website analysis and improving detection capabilities. The framework is designed to be applicable in the real world, considering factors such as effectiveness, speed of detection, scalability, adaptation, flexibility, and robustness. The study introduces a novel framework that evaluates the effectiveness, robustness, and detection speed of the hybrid approach. A proof of concept is developed to test these factors, demonstrating that the hybrid framework outperforms individual models in terms of accuracy, robustness, and detection speed. The framework uses a stacking function to combine predictions from different models, enhancing the overall performance. The results show that the hybrid framework achieves an accuracy of 97.44%, surpassing current state-of-the-art approaches while requiring less computational time. The study also introduces an innovative methodology for simulating bypass attacks on single-analysis models, highlighting the hybrid approach's resilience against sophisticated phishing attempts. The framework combines three deep learning models, each analyzing a different website feature: URL, HTML content, and HTML DOM tree structure. The URL-based model processes the URL as text, analyzing word and character relationships. The HTML content-based model analyzes the HTML code, while the HTML DOM tree analysis extracts the DOM structure and processes it as a text sequence. Each model makes predictions for the test set, which are then combined using stacking functions. The study assesses the effectiveness of the hybrid framework, its detection speed, and its robustness against bypassing attempts. The results indicate that the hybrid framework is more robust and effective than individual models, making it a promising solution for automated phishing detection. The framework provides a general basis for applicable and robust phishing detection architectures, contributing to the field of cybersecurity by advancing the understanding of hybrid approaches and their potential to enhance detection models. The study also highlights the importance of considering multiple factors of applicability in designing real-world phishing detection systems.This paper presents a hybrid framework for automated phishing detection, which combines multiple models to enhance both effectiveness and robustness. Phishing attacks are a critical cybersecurity threat, and current systems often rely on single-analysis models, making them vulnerable to sophisticated bypass attempts. The proposed hybrid framework addresses these limitations by integrating different models, allowing for more comprehensive website analysis and improving detection capabilities. The framework is designed to be applicable in the real world, considering factors such as effectiveness, speed of detection, scalability, adaptation, flexibility, and robustness. The study introduces a novel framework that evaluates the effectiveness, robustness, and detection speed of the hybrid approach. A proof of concept is developed to test these factors, demonstrating that the hybrid framework outperforms individual models in terms of accuracy, robustness, and detection speed. The framework uses a stacking function to combine predictions from different models, enhancing the overall performance. The results show that the hybrid framework achieves an accuracy of 97.44%, surpassing current state-of-the-art approaches while requiring less computational time. The study also introduces an innovative methodology for simulating bypass attacks on single-analysis models, highlighting the hybrid approach's resilience against sophisticated phishing attempts. The framework combines three deep learning models, each analyzing a different website feature: URL, HTML content, and HTML DOM tree structure. The URL-based model processes the URL as text, analyzing word and character relationships. The HTML content-based model analyzes the HTML code, while the HTML DOM tree analysis extracts the DOM structure and processes it as a text sequence. Each model makes predictions for the test set, which are then combined using stacking functions. The study assesses the effectiveness of the hybrid framework, its detection speed, and its robustness against bypassing attempts. The results indicate that the hybrid framework is more robust and effective than individual models, making it a promising solution for automated phishing detection. The framework provides a general basis for applicable and robust phishing detection architectures, contributing to the field of cybersecurity by advancing the understanding of hybrid approaches and their potential to enhance detection models. The study also highlights the importance of considering multiple factors of applicability in designing real-world phishing detection systems.