This meta-analysis of randomized controlled trials (published or unpublished RCTs in English) examined the effectiveness of dance interventions compared to waitlist/active controls on cognitive functions in patients with mild cognitive impairment (MCI). Literature search was conducted on MEDLINE, EMBASE, CINAHL and related databases through 3 August 2019. Risk of bias was assessed with the PEDro scale regarding randomization, allocation concealment, group comparability, blinding, attrition and selective reporting. Five included RCTs (N = 358) used Latin, ballroom and aerobic dances (frequency: 1-3 session/week; intensity: light to moderate; time: 25-60 min/session; type: aerobic). Results show that dance interventions improve global cognition (SMD [95 % CI] = 0.48 [0.21, 0.74]), attention (SMD [95 % CI] = 0.33 [0.12, 0.54]), immediate (SMD [95 % CI] = 0.54 [0.38, 0.71]) and delayed recall (SMD [95 % CI] = 0.33 [0.01, 0.64]), and visuospatial ability (SMD [95 % CI] = 0.16 [0.01, 0.32]). Study limitations include a small number of RCTs and high heterogeneity in some cognitive domains. In summary, dance interventions improve some cognitive domains in MCI. (PROSPERO registration: CRD42019145418).