Limited medication adherence and persistence with treatment are barriers to successful management of type 2 diabetes (T2D). We searched MEDLINE, EMBASE, the Cochrane Library, the Register of Controlled Trials, PsychINFO and CINAHL for observational and interventional studies that compared the adherence or persistence associated with 2 or more glucose‐lowering medications in people with T2D. Where 5 or more studies provided the same comparison, a random‐effects meta‐analysis was performed, reporting mean difference (MD) or odds ratio (OR) for adherence or persistence, depending on the pooled study outcomes. We included a total of 48 studies. Compared with metformin, adherence (%) was better for sulphonylureas (5 studies; MD 10.6%, 95% confidence interval [CI] 6.5‐14.7) and thiazolidinediones (TZDs; 6 studies; MD 11.3%, 95% CI 2.7%‐20.0%). Adherence to TZDs was marginally better than adherence to sulphonylureas (5 studies; MD 1.5%, 95% CI 0.1‐2.9). Dipeptidyl peptidase‐4 inhibitors had better adherence than sulphonylureas and TZDs. Glucagon‐like peptide‐1 receptor agonists had higher rates of discontinuation than long‐acting analogue insulins (6 studies; OR 1.95; 95% CI 1.17‐3.27). Long‐acting insulin analogues had better persistence than human insulins (5 studies; MD 43.1 days; 95% CI 22.0‐64.2). The methods used to define adherence and persistence were highly variable.