The restorative benefits of the natural environment are crucial for human well-being and sustainable development. Although stress reduction and attention restoration through natural exposure have been quantified through physiological and psychological pathways, numerous studies have intentionally constructed idealized natural settings devoid of individuals to minimize interference. This deliberate approach has raised concerns about the accuracy of these restorative results, as real-world settings invariably involve other people. To address this issue, we designed and executed a randomized controlled experiment. By measuring physiological and psychological indicators and utilizing a two-way ANOVA and Tukey post-hoc comparison, we explored the restorative potential of natural settings within seven distinct social distances across five landscape types in virtual reality environments. The results revealed that beyond a social distance of 3.8 m, the presence of people had minimal impact on stress reduction, with attention restoration effects remaining consistently positive. Optimal physiological and psychological restoration is achieved when the social distance exceeded 20 m. Further exploration is warranted to elucidate the influence of landscape types on the restoration of natural environments. The findings provide valuable insights for the planning and design of restorative natural settings, supporting research endeavors aimed at improving human health and well-being and allowing for sustainable management.