Introduction Physical activity (PA) is prospectively inversely associated with dementia risk, but few studies examined accelerometer measures of PA and sitting with rigorously‐adjudicated mild cognitive impairment (MCI) and dementia risk. Methods We examined the associations of accelerometer measures (PA and sitting) with incident MCI/probable dementia in the Women's Health Initiative (n = 1277; mean age = 82 ± 6 years) Results Over a median follow‐up of 4.2 years, 267 MCI/probable dementia cases were identified. Adjusted Cox regression HRs (95% CI) across moderate‐to‐vigorous PA (MVPA) min/d quartiles were 1.00 (reference), 1.28 (0.90 to 1.81), 0.79 (0.53 to 1.17), and 0.69 (0.45 to 1.06); P‐trend = 0.01. Adjusted HRs (95% CI) across steps/d quartiles were 1.00 (reference), 0.73 (0.51 to 1.03), 0.64 (0.43 to 0.94), and 0.38 (0.23 to 0.61); P‐trend < 0.001. The HR (95% CI) for each 1‐SD increment in MVPA (31 min/d) and steps/d (1865) were 0.79 (0.67 to 0.94) and 0.67 (0.54 to 0.82), respectively. Sitting was not associated with MCI/probable dementia. Discussion Findings suggest ≥ moderate intensity PA, particularly stepping, associates with lower MCI and dementia risk. Highlights Few studies have examined accelerometer‐measured physical activity, including steps, and sitting with incident ADRD. Moderate‐to‐vigorous physical activity and steps, but not light physical activity or sitting, were inversely associated with lower ADRD risk. Among older women, at least moderate intensity physical activity may be needed to reduce ADRD risk.