<p>As virtual reality (VR) becomes more widely adopted for instructional purposes, research also shows that VR-based learning can be emotionally involving and cognitively demanding, creating a need for instructional supports that respond to learners’ changing cognitive-emotional states. This two-group quasi-experimental quantitative study examined whether adaptively delivered proximity-framed prompts support emotional engagement during VR-based learning. Grounded in construal level theory, the prompts were designed to reduce psychological distance and increase immediate personal relevance at moments of disengagement. Seventy-seven volunteer undergraduate nursing students learned with a desktop VR-based medication administration simulation and were allocated by interval sampling to either adaptive proximity prompts (<i>n</i> = 35) or no prompts (<i>n</i> = 42). The system triggered prompts in real time using participant-specific indicators derived from facial expression analytics and blink-rate data. Emotional engagement was measured continuously using facial-expression measures, prompt responses were coded for psychological distance across social, temporal, spatial, and hypotheticality dimensions, and declarative and procedural knowledge were assessed with a knowledge test. Main analyses included linear mixed-effects models, Spearman correlations, and paired and independent-samples t tests. Both groups showed gains in declarative and procedural knowledge. Compared with the control condition, the adaptive-prompt condition showed shorter expression times for surprise and fear, with the reduction in surprise especially evident during procedural phases. Lower psychological distance in prompt responses was also associated with lower surprise and fear during procedural learning. Although no between-group differences emerged in immediate posttest performance, the findings show that adaptive proximity prompts can reshape the emotional profile of VR learning.</p>