This study investigated the effectivity and sustainability of a physical activity (PA) promotion and motor training programs and analyzed predictors for PA changes in persons with dementia. A total of 122 participants with mild-to-moderate dementia were randomized to the intervention program designed for persons with dementia (intervention group) or a motor placebo activity (control group). The primary outcome was the Physical Activity Questionnaire for the Elderly assessed at the baseline, after the 3-month intervention, and at a 3-month follow-up. The PA promotion program significantly increased PA in the intervention group compared with the control group during the training intervention phase. Both groups showed an increase in habitual PA when intervention-induced activities were excluded. PA was sustainably increased in both groups at follow-up. Low baseline PA was predictive for increased PA after the intervention and low baseline PA, high motor performance, and low comorbidity for increased PA at follow-up.