Abstract Aims/hypothesis We aimed to determine the longitudinal association of circulating markers of systemic inflammation with subsequent long-term cognitive change in older people with type 2 diabetes. Methods The Edinburgh Type 2 Diabetes Study is a prospective cohort study of 1066 adults aged 60 to 75 years with type 2 diabetes. Baseline data included C-reactive protein, IL-6, TNF-α fibrinogen and neuropsychological testing on major cognitive domains. Cognitive testing was repeated after 10 years in 581 participants. A general cognitive ability score was derived from the battery of seven individual cognitive tests using principal component analysis. Linear regression was used to determine longitudinal associations between baseline inflammatory markers and cognitive outcomes at follow-up, with baseline cognitive test results included as covariables to model cognitive change over time. Results Following adjustment for age, sex and baseline general cognitive ability, higher baseline fibrinogen and IL-6 were associated with greater decline in general cognitive ability (standardised βs = −0.059, p=0.032 and −0.064, p=0.018, respectively). These associations lost statistical significance after adjustment for baseline vascular and diabetes-related covariables. When assessing associations with individual cognitive tests, higher IL-6 was associated with greater decline in tests of executive function and abstract reasoning (standardised βs = 0.095, p=0.006 and −0.127, p=0.001, respectively). Similarly, raised fibrinogen and C-reactive protein levels were associated with greater decline in processing speed (standardised βs = −0.115, p=0.001 and −0.111, p=0.001, respectively). These associations remained statistically significant after adjustment for the diabetes- and vascular-related risk factors. Conclusions/interpretation Higher baseline levels of inflammatory markers, including plasma IL-6, fibrinogen and C-reactive protein, were associated with subsequent cognitive decline in older people with type 2 diabetes. At least some of this association appeared to be specific to certain cognitive domains and to be independent of vascular and diabetes-related risk factors. Graphical abstract