The current absence of a disease-modifying treatment for Alzheimer disease highlights the necessity for the benefits of nonpharmacological approaches. We aimed to investigate the effect of exercise in older patients with Alzheimer dementia.This is an observational, prospective cohort study in medical center. Eighty older patients with Alzheimer dementia, including 54 with mild dementia and 26 with moderate dementia, were followed up over 2 years. Patients were divided into exercise and no-exercise groups according to their weekly exercise habit. Mini-Mental State Examination (MMSE), clinical dementia rating (CDR), and senior fitness test were checked initially. We defined death and unexpected hospitalization as the outcomes.Age, sex, education years, and MMSE showed no significant differences between the groups (P > .05) in all patients. All the patients of the exercise group had significantly better left upper body strength, higher aerobic endurance, and left and right balance maintenance time than those of the no-exercise group (P < .05). There were no changes in hospitalization and mortality between the exercise and non-exercise groups during the 2-year follow-ups in all participants. However, in the mild and moderate dementia subgroups, age, sex, education years, and MMSE showed no significant differences between the groups (P > .05). The exercise group had significantly better lower body strength, left upper body strength, aerobic endurance, right upper body flexibility, lower body flexibility, balance maintenance, and agility than the no-exercise group in patients with mild dementia (P < .05). Moreover, the exercise group had significantly lesser unexpected hospitalization than the no-exercise group in the patients with mild dementia (P = .037).Despite the similarity in the status of dementia, exercise habit was found to be associated with a better senior fitness test score status. Hence, exercise can decrease unexpected hospitalization in patients with mild dementia but not those with total dementia.