Bronchoscopy is a minimally invasive procedure for diagnosing and treating lung conditions, but accurate navigation remains challenging and resource-intensive due to reliance on preoperative imaging, sensor-based tracking, and the low-saliency visual environment of the airways. To address these limitations, we propose a novel Navigational Bronchoscopy framework that enables real-time guidance and repeatable interventions without requiring external sensors or CT scans, making it particularly suitable for mechanically ventilated patients in critical care units with limited access to preoperative imaging. Our approach leverages deep learning, combining airway landmark recognition with deep visual features and a Vision Transformer (ViT)-based pose regression network to track bronchoscope motion. The framework is deployed on a commercially available bronchoscope and validated through trials in both a phantom lung model and a mechanically ventilated ex-vivo human lung. Results show that our ViT-based model achieves the lowest pose estimation errors among tested methods. Furthermore, in ex-vivo trials, our system successfully guided the bronchoscope to predefined targets, achieving high similarity scores for reliable landmark identification. These findings highlight the feasibility of our approach for real-world clinical applications.