Background and Purpose: Mild Cognitive Impairment (MCI) is associated with a greater risk of dementia for older adults. However, systematic reviews have shown that some physical exercise (PE) seems to improve MCI symptoms and signs. Those reviews and meta-analysis could not explain what possible moderator influenced their results. This meta-analysis aims to identify the effect of PE over older people's cognition with MCI and explore sources of heterogeneity. Methods: Databases were searched from inception January 2020 for randomized clinical trials that evaluated the effects of PE over cognition of older persons with MCI. Random effect meta-analyses were performed for each cognitive outcome. Subgroup analyses and meta-regressions models explored the potential sources of heterogeneity. Results: A total of 2077 participants (mean age = 71.8 years) from 27 studies were included. PE improves global cognitive function (SMD = 0.348 [95 % CI 0.166 to 0.529]; p = 0.0001), executive function (SMD = 0.213 [95 % CI 0.026 to 0.400]; p = 0.026) and delayed recall (SMD = 0.180 [95 % CI 0.002 to 0.358]; p = 0.047). A trend towards beneficial effects of PE on verbal fluency (SMD = 0.270 [95 %, CI -0.021 to 0.561]; p = 0.069) and attention (SMD = 0.170 [CI -0.016 to 0.357]; p = 0.073) were also observed. Subgroup analyses showed a relationship between modality and intensity of physical exercise and changes observed in global cognitive function, executive function, delayed recall, verbal fluency and working memory. Discussion and Conclusion: PE can ameliorate cognitive deficts of older adults with MCI. The most pronounced effects appear to arise from other types of exercise that included mind-body exercises and moderate intensity.