Dementia is a common medical condition in the ageing population, and cognitive intervention is a non-pharmacologic strategy to improve cognitive functions. This meta-analysis evaluated the benefits of computerized cognitive training (CCT) on memory functions in individuals with MCI or dementia. The study was registered prospectively with PROSPERO under CRD42022363715 and received no funding. The search was conducted on MEDLINE, Embase, and PsycINFO on Sept 19, 2022, and Google Scholar on May 9, 2023, to identify randomized controlled trials that examined the effects of CCT on memory outcomes in individuals with MCI or dementia. Mean differences and standard deviations of neuropsychological assessment scores were extracted to derive standardized mean differences. Our search identified 10,678 studies, of which 35 studies were included. Among 1489 participants with MCI, CCT showed improvements in verbal memory (SMD (95%CI) = 0.55 (0.35-0.74)), visual memory (0.36 (0.12-0.60)), and working memory (0.37 (0.10-0.64)). Supervised CCT showed improvements in verbal memory (0.72 (0.45-0.98)), visual memory (0.51 (0.22-0.79)), and working memory (0.33 (0.01-0.66)). Unsupervised CCT showed improvement in verbal memory (0.21 (0.04-0.38)) only. Among 371 participants with dementia, CCT showed improvement in verbal memory (0.64 (0.02-1.27)) only. Inconsistency due to heterogeneity (as indicated by I values) is observed, which reduces our confidence in MCI outcomes to a moderate level and dementia outcomes to a low level. The results suggest that CCT is efficacious on various memory domains in individuals with MCI. Although the supervised approach showed greater effects, the unsupervised approach can improve verbal memory while allowing users to receive CCT at home without engaging as many healthcare resources.