The prevalence of Alzheimer's disease (AD) is increasing, therefore, identifying biomarkers to predict those vulnerable to AD is imperative. Type 2 diabetes (T2D) serves as an independent risk factor for AD. Early prediction of T2D patients who may be more susceptible to AD, so as to achieve early intervention, is of great significance to reduce the prevalence of AD. To establish periphery biomarkers that could predict conversion of T2D into pre-AD-like cognitive decline. A follow-up study was carried out from 159 T2D patients at baseline. The correlations of cognitive states (by MMSE score) with multi-periphery biomarkers, including APOE genotype, plasma amyloid-β level, platelet GSK-3β activity, and olfactory score were analyzed by logistic regression. ROC curve was used for establishing the prediction model. Additionally, MRI acquired from 38 T2D patients for analyzing the correlation among cognitive function, biomarkers and brain structure. Compared with the patients who maintained normal cognitive functions during the follow-up period, the patients who developed MCI showed worse olfactory function, higher platelet GSK-3β activity, and higher plasma Aβ42/Aβ40 ratio. We conducted a predictive model which T2D patients had more chance of suffering from pre-AD-like cognitive decline. The MRI data revealed MMSE scores were positively correlated with brain structures. However, platelet GSK-3β activity was negatively correlated with brain structures. Elevated platelet GSK-3β activity and plasma Aβ42/Aβ40 ratio with reduced olfactory function are correlated with pre-AD-like cognitive decline in T2D patients, which used for predicting which T2D patients will convert into pre-AD-like cognitive decline in very early stage.