During the pandemic, visiting counts, dwell time, and travel distance to green space significantly declined. Previous studies have not fully explained why this pattern occurred, as they lack the simultaneous inclusion of both objective features and visitor’s experience, and do not explicitly distinguish visitation patterns into three categories. To examine how U.S. green space visitation patterns changed during the COVID-19 pandemic, I integrated Google reviews, Street View images, and the Social Vulnerability Index with GPS-based visitation data to examine how experience, environmental features, and socioeconomic sta- tus affected green space visitation, and further to identify features that fulfill basic and non-basic needs in different time periods. Machine learning models were used to transform raw data into numerical metrics and capture the non-linear impact of these features on green space visitation. The findings reveal that during the pandemic, basic needs such as green coverage and road are regarded as the most important. Non-basic features, such as water activities and sports, can still attract a smaller proportion of visitors, provided they have access to these green spaces.