Background Circular RNAs (circRNAs) have been demonstrated to be associated with Alzheimer's disease (AD). Here, we conducted a study to explore whether circRNAs have the ability to differentiate AD from cognitively normal controls and other types of dementia, such as vascular dementia (VaD), Parkinson's disease dementia (PDD), behavioral variant frontotemporal dementia (bvFTD), and dementia with Lewy body (DLB). Methods Three datasets were included in this study to measure blood circRNAs. The pilot study (Dataset 1, n = 40; controls, 20; AD, 20) was used to screen differentially expressed circRNAs. Dataset 2 (n = 124; controls, 61; AD, 63) was recruited for the establishment of the diagnostic model using a circRNA panel. Further, the Dataset 3 (n = 321; control, 58; AD, 60; VaD, 50; PDD, 51; bvFTD, 52; DLB, 50) was used to verify the diagnostic model. Results In Dataset 1, 22 upregulated and 19 downregulated circRNAs were revealed. In Dataset 2, a six-circRNA panel was found to be able to distinguish patients with AD from controls. Then this panel was applied to Dataset 3 and successfully differentiated AD from other types of dementia. Conclusion This study suggested that a six-circRNA panel is AD-specific and a promising biomarker of AD.