Long non-coding RNAs (lncRNAs) have been identified to be involved in the pathogenesis of Alzheimer's disease (AD). In this study, we evaluated whether lncRNAs can be used to discriminate AD patients from controls and patients with other dementias, such as vascular, Parkinson's disease, behavioral variant frontotemporal, and dementia with Lewy body. In this study, we used three datasets to measure the blood lncRNA levels. A pilot study (dataset 1, n = 40; controls, 20; AD, 20) was used to screen for differentially expressed lncRNAs. Dataset 2 (n = 174; controls, 86; AD, 88) was used to identify a lncRNA panel for the diagnostic model. Dataset 3 (n = 333; control, 60; AD, 54; vascular dementia, 53; Parkinson's disease dementia, 55; behavioral variant frontotemporal dementia, 56; and dementia with Lewy body, 55) was used to validate the diagnostic model. In dataset 1, 12 upregulated and 15 downregulated lncRNAs were identified. In dataset 2, a panel of seven lncRNAs was found to have the ability to differentiate AD patients from controls. Finally, this panel was applied to dataset 3 to successfully distinguish AD from other dementias. This study proposes a panel of seven lncRNAs as specific and promising biomarker for AD diagnosis.