OBJECTIVE: To evaluate the feasibility and safety of a home-based exercise programme for people with Alzheimer's disease, and to provide preliminary evidence of programme effectiveness in improving balance and mobility and reducing falls risk. DESIGN: A randomized controlled trial. SETTING: Community. PARTICIPANTS: Forty people with mild to moderate Alzheimer's disease (mean age 81.9, SD 5.72; 62.5% female). INTERVENTIONS: Participants were randomized to a six-month home-based individually tailored balance, strengthening and walking exercise programme (physiotherapist) or a six-month home-based education programme (control) (occupational therapist). Both programmes provided six home-visits and five follow-up phone calls. MAIN MEASURES: Balance, mobility, falls and falls risk were measured at baseline and programme completion. Intention-to-treat analysis using a generalized linear model with group allocation as a predictor variable was performed to evaluate programme effectiveness. Feasibility and adverse events were systematically recorded at each contact. RESULTS: Fifty-eight per cent of the exercise group finished the programme, completing an average of 83% of prescribed sessions, with no adverse events reported. Functional Reach improved significantly (P = 0.002) in the exercise group (mean (SD), 2.28 (4.36)) compared to the control group (-2.99 (4.87)). Significant improvement was also observed for the Falls Risk for Older People - Community score (P = 0.008) and trends for improvement on several other balance, mobility, falls and falls risk measures for the exercise group compared to the control group. CONCLUSIONS: The exercise programme was feasible and safe and may help improve balance and mobility performance and reduce falls risk in people with Alzheimer's disease.