Bright light treatment is an effective way to influence circadian rhythms in healthy adults, but previous research with dementia patients has yielded mixed results. The present study presents a primary outcome of the DEM.LIGHT trial, a 24-week randomized controlled trial conducted at nursing homes in Bergen, Norway, investigating the effects of a bright light intervention. The intervention consisted of ceiling-mounted LED panels providing varying illuminance and correlated color temperature throughout the day, with a peak of 1000 lx, 6000 K between 10 a.m. and 3 p.m. Activity was recorded using actigraphs at baseline and after 8, 16, and 24 weeks. Non-parametric indicators and extended cosine models were used to investigate rest–activity rhythms, and outcomes were analyzed with multi-level regression models. Sixty-one patients with severe dementia (median MMSE = 4) were included. After 16 weeks, the acrophase was advanced from baseline in the intervention group compared to the control group (B = −1.02, 95%; CI = −2.00, −0.05). There was no significant difference between the groups on any other rest–activity measures. When comparing parametric and non-parametric indicators of rest–activity rhythms, 25 out of 35 comparisons were significantly correlated. The present results indicate that ambient bright light treatment did not improve rest–activity rhythms for people with dementia.