Autoantibodies are self-antigen reactive antibodies that play diverse roles in the normal immune system, tissue homeostasis, and autoimmune and neurodegenerative diseases. Anti-neuronal autoantibodies have been detected in neurodegenerative disease serum, with unclear significance. To identify diagnostic biomarkers of Alzheimer's disease (AD), we analyzed serum autoantibody profiles of the HuProt proteome microarray using the discovery set of cognitively normal control (NC, n = 5) and AD (n = 5) subjects. Approximately 1.5-fold higher numbers of autoantibodies were detected in the AD group (98.0 ± 39.9/person) than the NC group (66.0 ± 39.6/person). Of the autoantigen candidates detected in the HuProt microarray, five autoantigens were finally selected for the ELISA-based validation experiment using the validation set including age- and gender-matched normal (NC, n = 44), mild cognitive impairment (MCI, n = 44) and AD (n = 44) subjects. The serum levels of four autoantibodies including anti-ATCAY, HIST1H3F, NME7 and PAIP2 IgG were significantly different among NC, MCI and/or AD groups. Specifically, the anti-ATCAY autoantibody level was significantly higher in the AD (p = 0.003) and MCI (p = 0.015) groups compared to the NC group. The anti-ATCAY autoantibody level was also significantly correlated with neuropsychological scores of MMSE (r = - 0.229, p = 0.012), K-MoCA (r = - 0.270, p = 0.003), and CDR scores (r = 0.218, p = 0.016). In addition, a single or combined occurrence frequency of anti-ATCAY and anti-PAIP2 autoantibodies was significantly associated with the risk of MCI and AD. This study indicates that anti-ATCAY and anti-PAIP2 autoantibodies could be a potential diagnostic biomarker of AD.