Recently, virtual environments (VEs) are suggested to encourage users to exercise regularly. The benefits of chronic exercise on cognitive performance are well documented in non-VE neurophysiological and behavioural studies. Based on event-related potentials (ERP) such as the N200 and P300, cognitive processing may be interpreted on a neuronal level. However, exercise-related neuroelectric adaptation in VE remains widely unclear and thus characterizes the primary aim of the present study. Twenty-two healthy participants performed active (moderate cycling exercise) and passive (no exercise) sessions in three VEs (control, front, surround), each generating a different sense of presence. Within sessions, conditions were randomly assigned, each lasting 5 min and including a choice reaction-time task to assess cognitive performance. According to the international 10:20 system, EEG with real-time triggered stimulus onset was recorded, and peaks of N200 and P300 components (amplitude, latency) were exported for analysis. Heart rate was recorded, and sense of presence assessed prior to and following each session and condition. Results revealed an increase in ERP amplitudes (N200: p < 0.001; P300: p < 0.001) and latencies (N200: p < 0.001) that were most pronounced over fronto-central and occipital electrode sites relative to an increased sense of presence (p < 0.001); however, ERP were not modulated by exercise (each p > 0.05). Hypothesized to mirror cognitive processing, decreases of cognitive performance's accuracy and reaction time failed significance. With respect to previous research, the present neuroelectric adaptation gives reason to believe in compensative neuronal resources that balance demanding cognitive processing in VE to avoid behavioural inefficiency.