Exercise is widely promoted for cognitive ageing, yet the field still lacks a clear account of what constitutes an effective “dose” when cognition is the target. From a movement-science perspective, this distinction is important because exercise interventions may differ not only in physical exposure, but also in the perceptual, attentional, coordinative and adaptive demands embedded in movement. Most trials define dose through physical exposure variables, including frequency, duration, weekly volume, intervention length and intensity. However, these metrics do not capture what the exercise task requires cognitively or motorically. We conducted a systematic review and meta-analysis of randomized controlled trials testing structured exercise in community-dwelling older adults without dementia. Cognitive outcomes were grouped into global cognition and domain-specific outcomes. Hedges’ g was calculated from post-intervention means and standard deviations, with positive values favoring exercise. Random-effects and multilevel models estimated pooled effects and tested whether weekly volume, intervention duration and a structured cognitive–motor demand score explained between-study variability. Thirty-seven trials provided 166 effect-size estimates. Structured exercise was associated with better global cognition (g = 0.55, 95% CI 0.38 to 0.73), although heterogeneity was substantial and certainty was very low. Domain-specific effects were smaller and generally favorable, but none remained significant after false discovery rate correction. Weekly volume and intervention duration did not consistently explain cognitive effects. Cognitive–motor demand showed only partial overlap with physical-dose indicators, suggesting that it captured a distinct qualitative dimension of exercise content; however, it did not explain effects beyond physical dose. Component-level analyses suggested possible domain-specific patterns for set shifting. Current trials are therefore not yet able to define a precise cognitive dose of exercise. Future studies should test physical dose and cognitive–motor task demands as distinct, potentially interacting dimensions of exercise for cognitive ageing.Systematic review registrationhttps://www.crd.york.ac.uk/PROSPERO/view/CRD420250622607.