Background: Ambient light therapies are potentially effective in improving sleep disturbances and circadian rhythms. This study created a new lighting intervention model for elderly patients with dementia. It is hypothesized that exposure to bright ambient light in the morning is more effective than general lighting in improving sleep disturbances and circadian rhythms. Methods: A single-blind longitudinal-group experimental design was employed. The dementia participants came from the community and nursing homes. Those in the experimental group were exposed to ambient light at 2500 lux, and those in the comparison group were exposed to 114–307 lux. The corresponding sleep disturbances and circadian rhythms were determined using an accelerometer (XA-5). A longitudinal experimental design was adopted to observe the time to an effective response. Results: The covariates of benzodiazepine use and total activity during the day were analyzed using generalized estimating equations. The experimental group showed significant sleep efficiency improvement, with mean increases of 41.9% (P < 0.001) and 31.7% (P = 0.002), sleep time increases of 141 min (P = 0.001) and 135 min (P = 0.008), awakening time decreases of 116 min (P = 0.001) and 108 min (P = 0.002), and sleep onset advancements of 60–84 min/sleep offset delays of 57–79 min upon the fifth and ninth week, respectively. A 4-week bright ambient light intervention was the most effective. Conclusions: This study found that bright ambient light in the morning is beneficial for improving sleep disturbances and was driven by stabilizing circadian rhythms.