Objectives: The robotic seal, PARO, has been used as an alternative to animal-assisted therapies with residents with dementia in long-term care, yet understanding of its efficacy is limited by a paucity of research. We explored the effects of PARO on motor activity and sleep patterns, as measured by a wearable triaxial accelerometer. Study design: Cluster-randomised controlled trial, involving 28 facilities in Queensland, Australia. Nine facilities were randomised to the PARO group (individual, non-facilitated, 15-min sessions three afternoons per week for 10 weeks), 10 to a plush toy (PARO with robotic features disabled) and nine to usual care. Main outcome measures: Changes in day- and nighttime motor activity and sleep after the 10-week intervention, as measured by SenseWear® armbands, worn by participants continuously for 24 h at baseline, during two single intervention days in weeks 5 and 10 respectively, and post-intervention (week 15). Analyses followed intention-to-treat, using repeated-measures mixed-effects models. Results: After 10 weeks, the PARO group showed a greater reduction in daytime step count than usual care (p = 0.023), and in nighttime step count (p = 0.028) and daytime physical activity (p = 0.026) compared with the plush toy group. At post-intervention, the PARO group showed a greater reduction in daytime step count than the plush toy group (p = 0.028), and at nighttime compared with both the plush toy group (p = 0.019) and the usual-care group (p = 0.046). The PARO group also had a greater reduction in nighttime physical activity than the usual-care group (p = 0.015). Conclusions: PARO may have some effect on motor activity of older people with dementia in long-term care, but not on sleep patterns. Australian New Zealand Clinical Trials Registry (ACTRN12614000508673).