Goal-oriented navigation is widely understood to depend critically upon internal maps. Although this may be the case in many settings, many animals, such as humans, tend to rely on vision when the environment is complex or unfamiliar. The nature of gaze during visually-guided navigation remains unknown. To study this, we tasked humans to navigate to transiently visible goals in virtual mazes of varying complexity, observing that they achieved near-optimal path lengths in all arenas. By analyzing subjects’ eye movements, we gained insights into the principles of active sensing and prospection that would not have been gleaned from observing navigational behavior alone. The spatial distribution of fixations revealed that environmental complexity mediated a striking trade-off in the extent to which attention was directed towards two complimentary aspects of the world model: the reward location and task-relevant transitions. Furthermore, the temporal evolution of gaze revealed rapid, sequential prospection of the future path, evocative of neural ’preplay’. These findings suggest that the spatiotemporal characteristics of eye movements during navigation are significantly shaped by the unique cognitive computations underlying real-world, sequential decision making.