Abstract Background dance is a mind–body activity that stimulates neuroplasticity. We explored the effect of dance on cognitive function in older adults. Methods we searched MEDLINE, EMBASE, CENTRAL and PsycInfo databases from inception to August 2020 (PROSPERO:CRD42017057138). Inclusion criteria were (i) randomised controlled trials (ii) older adults (aged ≥ 55 years), (iii) intervention—dance and (iv) outcome—cognitive function. Cognitive domains were classified with the Diagnostic and Statistical Manual of Mental Disorders-5 Neurocognitive Framework. Meta-analyses were performed in RevMan5.3 and certainty of evidence with GradePro. Results we reviewed 3,997 records and included 11 studies (N = 1,412 participants). Seven studies included only healthy older adults and four included those with mild cognitive impairment (MCI). Dance interventions varied in frequency (1–3×/week), time (35–60 minutes), duration (3–12 months) and type. We found a mean difference (MD) = 1.58 (95% confidence interval [CI) = 0.21–2.95) on the Mini Mental State Examination for global cognitive function (moderate-certainty evidence), and the Wechsler Memory Test for learning and memory had an MD = 3.02 (95% CI = 1.38–4.65; low-certainty evidence). On the Trail Making Test-A for complex attention, MD = 3.07 (95% CI = −0.81 to 6.95; high-certainty evidence) and on the Trail Making Test-B for executive function, MD = −4.12 (95% CI = −21.28 to 13.03; moderate-certainty evidence). Subgroup analyses did not suggest consistently greater effects in older adults with MCI. Evidence is uncertain for language, and no studies evaluated social cognition or perceptual–motor function. Conclusions dance probably improves global cognitive function and executive function. However, there is little difference in complex attention, and evidence also suggests little effect on learning and memory. Future research is needed to determine the optimal dose and if dance results in greater cognitive benefits than other types of physical activity and exercise.