The gaze pattern changes in old age, not only during artificial laboratory tasks but also during quasi-natural behavior. We have recently reported that older adults, walking in a virtual reality pedestrian precinct, spent longer time looking at pedestrian traffic lights than young adults did (Bock et al, 2015). We have interpreted this age-related change as a compensatory strategy, and we now analyze whether this strategy might be potentially hazardous in that it withdraws gaze from other regions that are critical for safe walking. Seventeen young and 16 older adults walked on a non-motorized treadmill linked to the 3D model of a pedestrian precint. The model was displayed on a monitor ahead, such that participants felt as if walking through the simulated world. Along their way, participants met a range of familiar objects such as pedestrian traffic lights, oncoming pedestrians and cats crossing their path. Eye position was recorded by a video-based system. We found that compared to young adults, older ones looked longer at regions of high behavioral relevance and less long at regions of low behavioral relevance. We conclude that looking longer at relevant regions might be a strategy for compensating central processing deficits, but this strategy may not pay off when an unexpected threat emerges in a seemingly irrelevant region.