This study examined users’ visual perception characteristics during the wayfinding process in underground complex spaces. The primary objective was to compare and analyze these characteristics based on users’ wayfinding abilities in both virtual reality (VR) and real-world environments, with the aim of identifying design strategies for effective wayfinding in underground spaces. The study was conducted at Jamsil Station’s undergrounㅔ가d complex in Seoul, involving 22 participants (average age = 26.5 years) who were classified by wayfinding ability following a preliminary survey (N = 30). Participants completed eye-tracking experiments in both VR and real-world settings. In the VR environment, participants wore headsets and navigated a designated route, which they later explored in the real space. The collected gaze data, including sequence, fixation time, and duration, were statistically analyzed using gaze data analysis and t-tests to examine differences based on wayfinding abilities. The findings revealed that participants with prior VR wayfinding experience processed visual information more efficiently in real spaces. In the VR environment, participants concentrated more on upper signboards and walls, focusing on specific visual cues, while in real spaces, their gaze was more dispersed as they utilized a broader range of visual information. The study concluded that prior learning in a virtual setting enhanced wayfinding performance in real environments. Additionally, the analysis of gaze patterns highlighted that visual information usage varied at complex nodal points based on wayfinding abilities, emphasizing the importance of user-centered information design in spatial planning.