Abstract Eye tracking is a promising research tool for investigating learning processes in immersive virtual reality (IVR). Grounded in cognitive theories of learning, this study proposes a classification of six key learning processes (i.e., cognitive, motivational, affective, social, metacognitive, media-specific) and nine associated constructs (i.e., attention, visual strategy, cognitive load, motivation, emotion, social attention, self-regulation, immersion, usability) measurable through eye tracking in IVR. Using this classification, a systematic review of 38 empirical studies published over a 10-year period (2014 − 2024) was conducted to evaluate the validity of eye-tracking measures. The results showed that the overall validity remains limited with an average score of 48% (± 19%), mainly due to insufficient control of data quality. Measures of cognitive load demonstrated relatively higher validity, likely due to their stronger alignment with performance outcomes. Methodological limitations, such as small sample sizes and dependence on default equipment configurations, were prevalent. To improve validity, future study should strengthen methodological foundations, provide detailed documentation of experimental setups, and use triangulation with complementary data. The classification and validity scoring schemes presented in this study contribute to the theory-driven application of eye tracking in IVR, and offer guidance for improving measurement validity in educational psychology.