Background and Purpose: Both adherence to the Mediterranean diet (MedDiet) and the use of metformin could benefit the cognitive performance of individuals with type 2 diabetes, but evidence is still controversial. We examined the association between metformin use and cognition in older adults with type 2 diabetes following a MedDiet intervention.Methods: Prospective cohort study framed in the PREDIMED-Plus-Cognition sub-study. The PREDIMED-Plus clinical trial aims to compare the cardiovascular effect of two MedDiet interventions, with and without energy restriction, in individuals with overweight/obesity and metabolic syndrome. The present sub-study included 487 cognitively normal subjects (50.5% women, mean ± SD age of 65.2 ± 4.7 years), 30.4% of them (N = 148) with type 2 diabetes. A comprehensive battery of neurocognitive tests was administered at baseline and after 1 and 3 years. Individuals with type 2 diabetes that exhibited a good glycemic control trajectory, either using or not using metformin, were compared to one another and to individuals without diabetes using mixed-effects models with inverse probability of treatment weights.Results: Most subjects with type 2 diabetes (83.1%) presented a good and stable glycemic control trajectory. Before engaging in the MedDiet intervention, subjects using metformin scored higher in executive functions (Cohen's d = 0.51), memory (Cohen's d = 0.38) and global cognition (Cohen's d = 0.48) than those not using metformin. However, these differences were not sustained during the 3 years of follow-up, as individuals not using metformin experienced greater improvements in memory (β = 0.38 vs. β = 0.10, P = 0.036), executive functions (β = 0.36 vs. β = 0.02, P = 0.005) and global cognition (β = 0.29 vs. β = −0.02, P = 0.001) that combined with a higher MedDiet adherence (12.6 vs. 11.5 points, P = 0.031). Finally, subjects without diabetes presented greater improvements in memory than subjects with diabetes irrespective of their exposure to metformin (β = 0.55 vs. β = 0.10, P < 0.001). However, subjects with diabetes not using metformin, compared to subjects without diabetes, presented greater improvements in executive functions (β = 0.33 vs. β = 0.08, P = 0.032) and displayed a higher MedDiet adherence (12.6 points vs. 11.6 points, P = 0.046).Conclusions: Although both metformin and MedDiet interventions are good candidates for future cognitive decline preventive studies, a higher adherence to the MedDiet could even outweigh the potential neuroprotective effects of metformin in subjects with diabetes.