Introduction The rapid rise in the prevalence of diabetes has a negative impact on patients' quality of life. Diabetes self-management group education is cost-effective and efficient for patients to control blood glucose. However, there are no consistent standards for self-management group education, and its long-term effects (≥12 months) are unclear. Although a few systematic reviews evaluated the long-term effects, they did not make clear provisions on the content of self-management, and the number and sample size of included studies were small, which may lead to misclassification bias and reporting bias. Therefore, we plan to conduct this systematic review to evaluate the long-term effects of self-management group education and determine the effects of different self-management characteristics on glycosylated haemoglobin (HbA1c). Methods and analysis We will retrieve Chinese databases (Wanfang, Chinese Hospital Knowledge Warehouse) and English databases (PubMed, ScienceDirect, EMBASE, Web of Science, Bailian Platform, Cochrane Central Register of Controlled Trials, Google Scholar) for randomly controlled trials and cluster randomly controlled trials of which participants are adults with type 2 diabetes mellitus. We will manually search citation lists and trial registries, and consult authors to obtain relevant articles. The retrieval time range will be from the establishment of the database to July 2020 to avoid omitting relevant studies. The primary outcome will be HbA1c. The secondary outcomes will be fasting plasma glucose, postprandial blood glucose, total cholesterol, triglyceride, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, systolic blood pressure, diastolic blood pressure, body mass index, waist circumference and death event. Two reviewers will independently conduct article screening and assessment of risk of bias, with a third reviewer arbitrating if necessary. We will give priority to the use of meta-analysis to evaluate the pooled effects of all outcomes. For the outcomes of unrecognised sources of heterogeneity, missing data and less than three related studies, narrative synthesis approach will be used. Ethics and dissemination Ethical approval is not required for this systematic review. We plan to present the findings in a peer-reviewed scientific journal, relevant and responsible organisations, and training meetings. PROSPERO registration number CRD42020209011.