Comprehensive Geriatric Assessment (CGA) is a multidimensional interdisciplinary diagnostic process focused on determining an older person's medical, psychological and functional capability in order to develop a co-ordinated and integrated care plan. CGA is not limited simply to assessment, but also directs a holistic management plan for older people, which leads to tangible interventions. While there is established evidence that CGA reduces the likelihood of death and disability in acutely unwell older people, the effectiveness of CGA for community-dwelling, frail, older people at risk of poor health outcomes is less clear. To determine the effectiveness of CGA for community-dwelling, frail, older adults at risk of poor health outcomes in terms of mortality, nursing home admission, hospital admission, emergency department visits, serious adverse events, functional status, quality of life and resource use, when compared to usual care. We searched CENTRAL, MEDLINE, Embase, CINAHL, three trials registers (WHO ICTRP, ClinicalTrials.gov and McMaster Aging Portal) and grey literature up to April 2020; we also checked reference lists and contacted study authors. We included randomised trials that compared CGA for community-dwelling, frail, older people at risk of poor healthcare outcomes to usual care in the community. Older people were defined as 'at risk' either by being frail or having another risk factor associated with poor health outcomes. Frailty was defined as a vulnerability to sudden health state changes triggered by relatively minor stressor events, placing the individual at risk of poor health outcomes, and was measured using objective screening tools. Primary outcomes of interest were death, nursing home admission, unplanned hospital admission, emergency department visits and serious adverse events. CGA was delivered by a team with specific gerontological training/expertise in the participant's home (domiciliary Comprehensive Geriatric Assessment (dCGA)) or other sites such as a general practice or community clinic (community Comprehensive Geriatric Assessment (cCGA)). Two review authors independently extracted study characteristics (methods, participants, intervention, outcomes, notes) using standardised data collection forms adapted from the Cochrane Effective Practice and Organisation of Care (EPOC) data collection form. Two review authors independently assessed the risk of bias for each included study and used the GRADE approach to assess the certainty of evidence for outcomes of interest. We included 21 studies involving 7893 participants across 10 countries and four continents. Regarding selection bias, 12/21 studies used random sequence generation, while 9/21 used allocation concealment. In terms of performance bias, none of the studies were able to blind participants and personnel due to the nature of the intervention, while 14/21 had a blinded outcome assessment. Eighteen studies were at low risk of attrition bias, and risk of reporting bias was low in 7/21 studies. Fourteen studies were at low risk of bias in terms of differences of baseline characteristics. Three studies were at low risk of bias across all domains (accepting that it was not possible to blind participants and personnel to the intervention). CGA probably leads to little or no difference in mortality during a median follow-up of 12 months (risk ratio (RR) 0.88, 95% confidence interval (CI) 0.76 to 1.02; 18 studies, 7151 participants (adjusted for clustering); moderate-certainty evidence). CGA results in little or no difference in nursing home admissions during a median follow-up of 12 months (RR 0.93, 95% CI 0.76 to 1.14; 13 studies, 4206 participants (adjusted for clustering); high-certainty evidence). CGA may decrease the risk of unplanned hospital admissions during a median follow-up of 14 months (RR 0.83, 95% CI 0.70 to 0.99; 6 studies, 1716 participants (adjusted for clustering); low-certainty evidence). The effect of CGA on emergency department visits is uncertain and evidence was very low certainty (RR 0.65, 95% CI 0.26 to 1.59; 3 studies, 873 participants (adjusted for clustering)). Only two studies (1380 participants; adjusted for clustering) reported serious adverse events (falls) with no impact on the risk; however, evidence was very low certainty (RR 0.82, 95% CI 0.58 to 1.17). CGA had no impact on death or nursing home admission. There is low-certainty evidence that community-dwelling, frail, older people who undergo CGA may have a reduced risk of unplanned hospital admission. Further studies examining the effect of CGA on emergency department visits and change in function and quality of life using standardised assessments are required.