Abstract Introduction To investigate how frailty status affects the outcome of exercise intervention among home-dwelling participants with Alzheimer disease (AD). Methods This is a sub-group analysis of a randomized controlled trial. In this trial, home-dwelling participants with AD received either home-based or group-based exercise twice a week for one year ( n = 129); the control group received normal care ( n = 65). Both the intervention and control group were subdivided into two groups according to modified Fried criteria: prefrail (0–1 criteria) and advanced frailty (2–5 criteria). The Functional Independence Measure (FIM) and number of falls per person-years served as outcome measures. Results Whereas there was no significant difference in FIM between the prefrail intervention (PRI) and control (PRC) groups at 3 or 6 months, the PRI group deteriorated significantly slower at 12 months (−6.6 95% CI −8.6 to −4.5 for PRI and −11.1 95% CI −13.9 to −8.3 for PRC; P = 0.010). Similarly, there was no significant difference between the advanced frailty intervention (AFI) and control (AFC) groups at 3 months, but the difference became significant at 6 months (−8.1 95% CI −11.1 to −5.2 for AFI and −15.5 95% CI −20.0 to −11.1 for AFC; P = 0.007) and at 12 months (−8.9 95% CI −11.9 to −5.9 for AFI and −15.3 95% CI −20.2 to −10.3 for AFC; P = 0.031). There was also a significant difference in the number of falls in favor of PRI and AFI groups compared to their respective control groups. Conclusion A long-term exercise intervention benefited people with AD regardless of their stage of frailty. Trial registration : ACTRN12608000037303.