Virtual reality (VR) offers compelling opportunities for human augmentation, with free three-dimensional (3D) flight enabling fluid navigation and immersive spatial exploration. However, most existing VR flight techniques rely either on abstract input methods—leading to sensorimotor mismatch and limited natural interaction—or on static embodied mappings that cannot adapt to the dynamic nature of movement. We introduce AdaptiveFly, an embodied flight technique based on bio-congruent adaptation. AdaptiveFly dynamically modulates the coupling between head rotation and yaw control according to user speed—loosening it at low speeds to allow free gaze movement, and tightening it at high speeds for stable steering. It also calibrates control sensitivity to individual posture characteristics and incorporates asymmetric pitch mapping to enhance the intuitiveness of vertical motion. A user study confirms that AdaptiveFly enhances naturalness, comfort, and coordination compared to both abstract and static embodied baselines. A proof-of-concept deployment on a physical drone further demonstrates the technique’s real-world applicability. This study expands the design of embodied VR locomotion and demonstrates the potential value of adaptive, human-consistent control strategies in 3D interfaces and human augmentation technologies.