This study aims to assess the effects of exercise on cognitive impairment behavioral performance and neuroprotective mechanisms in diabetes mellitus (DM) animal models. PubMed, Embase, Web of Science, China National Knowledge Infrastructure (CNKI), Wanfang Database, VIP Database (VIP), and China Biomedical Literature Database (CBM) were systematically searched for studies investigating the impact of exercise on cognitive impairment in animal models of diabetes mellitus (DM) from the inception of these databases through July 2023. Rigorous quality assessments were conducted on the included literature. Primary outcome measures comprised fasting blood glucose (FBG) levels and performance in the Morris water maze test, while secondary outcomes focused on mechanisms related to neuroprotection. Statistical analysis of outcome data was conducted using RevMan 5.3 and R software. A total of 17 studies were included, encompassing 399 animals. The results of the meta-analysis of primary outcome measures revealed that, compared to the control group, exercise effectively reduced fasting blood glucose (FBG) levels in diabetic animal models. In the Morris water maze experiment, exercise also significantly decreased the escape latency of diabetic animal models, increased the number of platform crossings, improved the percentage of time spent in the target quadrant, extended the time spent in the target quadrant, and enhanced swimming speed. Meta-analysis of secondary outcome measures indicated that exercise effectively reduced Aβ deposition, attenuated oxidative stress, enhanced synaptic function, suppressed cellular apoptosis and neuroinflammation, and promoted neurogenesis. Exercise represents a promising non-pharmacological therapy with a positive impact on diabetes-related cognitive function and neuroprotection. Moreover, this study provides a theoretical foundation for further preclinical and clinical trials.