This study aimed to investigate the controversial association between metformin use and diabetes-associated dementia in elderly patients with type 2 diabetes mellitus (T2DM) and evaluate the potential protective effects of metformin, as well as its intensity of use and dose-dependency, against dementia in this population. The study used a time-dependent Cox hazards model to evaluate the effect of metformin use on the incidence of dementia. The case group included elderly patients with T2DM (≥60 years old) who received metformin, while the control group consisted of elderly patients with T2DM who did not receive metformin during the follow-up period. Our analysis revealed a significant reduction in the risk of dementia among elderly individuals using metformin, with an adjusted hazard ratio of 0.34 (95% confidence interval: 0.33 to 0.36). Notably, metformin users with a daily intensity of 1 defined daily dose (DDD) or higher had a lower risk of dementia, with an adjusted hazard ratio (95% confidence interval) of 0.46 (0.22 to 0.6), compared to those with a daily intensity of <1 DDD. Additionally, the analysis of cumulative DDDs of metformin showed a dose-response relationship, with progressively lower adjusted hazard ratio across quartiles (0.15, 0.21, 0.28, and 0.53 for quartiles 4, 3, 2 and 1, respectively), compared to never metformin users (P for trend < 0.0001). Metformin use in elderly patients with T2DM is significantly associated with a substantial reduction in the risk of dementia. Notably, the protective effect of metformin demonstrates a dose-dependent relationship, with higher daily and cumulative dosages of metformin showing a greater risk reduction.