Background Age-related cognitive decline is a pervasive problem in the ageing population. Baduanjin training is a mind-body exercise with the characteristics of traditional Chinese medicine, and increasing numbers of studies have reported its usefulness in modulating the cognitive performance of various populations. However, no systematic review has evaluated the effect of Baduanjin training on cognition in middle-aged and older adults. Objective To systematically evaluate the effects of Baduanjin on the global cognitive function and specific cognitive domains of middle-aged and elderly people. Methods Four literature databases (PubMed, Cochrane library, EMBASE, and Web of Science) and four Chinese databases (Wanfang, China National Knowledge Infrastructure, Chinese Science and Technology Periodical and China Biology Medicine) were searched from inception through May 2020. Randomized controlled trials (RCTs) examining the effect of Baduanjin exercise on the cognitive function of middle-aged and elderly people were included. Assessment of the risk of bias for the included studies and data synthesis were conducted using the software Review Manager 5.3 based on the methods given in the Cochrane Handbook. Results Baduanjin training showed significant benefit for global cognitive function and parts of specific domains of cognition, including general memory and its sub-domains (i.e., immediate memory and delayed memory), executive function, and processing speed, but no significant difference was found in attention function, visual-spatial ability or long-term memory (a sub-domain of memory). No related adverse events were reported in the included studies. Conclusions The findings of this review suggest that Baduanjin is safe and effective in enhancing global cognitive function and memory in middle-aged and older adults and potentially beneficial to parts of the other specific domains of cognition, including executive function and processing speed. However, additional trials with larger sample sizes and a more rigorous design are needed before more definitive conclusions can be drawn.Abbreviations AVLTauditory language learning testCDTclock drawing testCIconfidence intervalDSCdigital symbol conversionFEMfixed-effect modelGo/No GoGo/No Go response test and correct-number testLOTCAloewenstein occupational therapy cognitive assessmentMCImild cognitive impairmentMDmean differenceMMSEmini-mental state examinationMoCAmontreal cognitive assessmentPFCprefrontal cortexRCTrandomized controlled trialREMrandom-effect modelrsFCresting-state Functional connectivitySMDstandardized mean differenceTMTtrail making testWMSwechsler memory scale