Frailty may represent a modifiable risk factor for dementia, but the direction of that association remains uncertain. We investigated frailty trajectories in the years preceding dementia onset using data from 23,672 participants (242,760 person-years of follow-up, 2,906 cases of incident dementia) across four cohort studies in the United States and United Kingdom. Bayesian non-linear models revealed accelerations in frailty trajectories 4-9 years before incident dementia. Among participants whose time between frailty measurement and incident dementia exceeded that prodromal period, frailty remained positively associated with dementia risk (adjusted hazard ratios ranged from 1.20 [95% confidence interval, CI = 1.15-1.26] to 1.43 [95% CI = 1.14-1.81]). This observational evidence suggests that frailty increases dementia risk independently of any reverse causality. These findings indicate that frailty measurements can be used to identify high-risk population groups for preferential enrolment into clinical trials for dementia prevention and treatment. Frailty itself may represent a useful upstream target for behavioural and societal approaches to dementia prevention.