Objectives. To determine the effects on balance and gait of a Wii-Fit program compared to a walking program in subjects with mild Alzheimer's dementia (AD). Methods. A prospective randomized (1 : 1) pilot study with two intervention arms was conducted in an assisted living facility with twenty-two mild AD subjects. In both groups the intervention occurred under supervision for 30 minutes daily, five times a week for eight weeks. Repeated measures ANOVA and paired t-tests were used to analyze changes. Results. Both groups showed improvement in Berg Balance Scale (BBS), Tinetti Test (TT) and Timed Up and Go (TUG) over 8 weeks. However, there was no statistically significant difference between the groups over time. Intragroup analysis in the Wii-Fit group showed significant improvement on BBS (P = 0.003), and TT (P = 0.013). The walking group showed a trend towards improvement on BBS (P = 0.06) and TUG (P = 0.07) and significant improvement in TT (P = 0.06). Conclusion. This pilot study demonstrates the safety and efficacy of Wii-Fit in an assisted living facility in subjects with mild AD. Use of Wii-Fit resulted in significant improvements in balance and gait comparable to those in the robust monitored walking program. These results need to be confirmed in a larger, methodologically sound study.