Generative AI as a cognitive co-learner: a developmental framework for AI literacy in health sciences education

Generative artificial intelligence (AI), especially large language models, is playing an expanding role in shaping learning within health sciences education. Current discussions often focus on efficiency or academic integrity, with less attention to how learners engage with AI across evolving cognitive and developmental stages. This Perspective conceptualizes generative AI as a cognitive co-learner, an interactive system that supports idea generation, organization, and reasoning while requiring human oversight for validation and interpretation. We propose a developmental framework for AI literacy that describes stage-typical patterns of AI engagement across educational contexts characterized by differing cognitive and epistemic demands. The proposed progression from exploration to synthesis to calibration reflects changes in the functional role of AI, learner cognitive engagement, and trust in AI-generated outputs. This progression is intended as a theoretical proposition rather than an empirically validated development trajectory. Central to this framework is calibrated trust, defined as the alignment between confidence in AI outputs and their actual reliability. Across all stages, human judgment, critical evaluation, and metacognitive awareness remain essential. This Perspective highlights the need for stage-appropriate educational strategies and learning activities that support responsible AI use while strengthening health sciences education and maintaining the primacy of human cognition.