Background: Structured activities to stimulate brain function—that is, cognitive training exercises—are promoted to slow or prevent cognitive decline, including dementia, but their effectiveness is highly debated. Purpose: To summarize evidence on the effects of cognitive training on cognitive performance and incident dementia outcomes for adults with normal cognition or mild cognitive impairment (MCI). Data Sources: Ovid MEDLINE, EMBASE, the Cochrane Central Register of Controlled Trials, and PsycINFO through July 2017, supplemented by hand-searches. Study Selection: Trials (published in English) lasting at least 6 months that compared cognitive training with usual care, waitlist, information, or attention controls in adults without dementia. Data Extraction: Single-reviewer extraction of study characteristics confirmed by a second reviewer; dual-reviewer risk-of-bias assessment; consensus determination of strength of evidence. Only studies with low or medium risk of bias were analyzed. Data Synthesis: Of 11 trials with low or medium risk of bias, 6 enrolled healthy adults with normal cognition and 5 enrolled adults with MCI. Trainings for healthy older adults were mostly computer based; those for adults with MCI were mostly held in group sessions. The MCI trials used attention controls more often than trials with healthy populations. For healthy older adults, training improved cognitive performance in the domain trained but not in other domains (moderate-strength evidence). Results for populations with MCI suggested no effect of training on performance (low-strength and insufficient evidence). Evidence for prevention of cognitive decline or dementia was insufficient. Adverse events were not reported. Limitation: Heterogeneous interventions and outcome measures; outcomes that mostly assessed test performance rather than global function or dementia diagnosis; potential publication bias. Conclusion: In older adults with normal cognition, training improves cognitive performance in the domain trained. Evidence regarding prevention or delay of cognitive decline or dementia is insufficient.