In order to examine the nonlinear effects of origin-based environmental features on the travel distance of older adults, Shanghai was selected as the case study. Mobile phone signaling data, street-view images, and multi-source built-environment data were used to investigate the relationships between environmental characteristics and travel behavior. Subjective perception evaluation results of street-view images were obtained using a Qwen-series multimodal large language model. A random forest model was used for nonlinear analysis. The results show that the density of daily service facilities, the density of medical facilities, and road density make large contributions. Their relative importance values are 17.70%, 14.81%, and 12.33%, respectively. These variables exhibit nonlinear and threshold effects on the travel distance of older adults. It is concluded that the density of daily service facilities and the density of medical facilities show negative response patterns. Their effects stabilize after about 45 facilities/km² and 5 facilities/km², respectively. Road density, green view index, and perceived interestingness show positive response patterns. Their thresholds are about 16 km/km², 23%, and 4.3, respectively. Perceived comfort shows a negative response pattern. Its effect gradually stabilizes after about 5.2. The findings provide theoretical support for urban spatial planning and the improvement of travel services for older adults.