© 2018 Societas Neurologica Japonica. All rights reserved. The epilepsy syndrome of autoimmune etiology, namely, autoimmune epilepsy has attracted attention in recent years, as was reflected in the new etiology of “immunity” in the Epilepsy Classification of the International League Against Epilepsy (2017). However, no specific tests other than neuronal antibodies have been established. We proposed a diagnostic algorithm for autoimmune epilepsy and preliminarily investigated its clinical utility. We applied this algorithm to 70 patients who were suspected as having autoimmune epilepsy from clinical symptoms and laboratory findings in our institute. At least one of the three neuronal antibodies (antibodies to N-methyl-D-aspartic acid receptor (NMDAR), glutamic acid decarboxylase (GAD), and voltage-gated potassium channels (VGKC) complex) was evaluated. In this two-step algorithm, the patients were initially screened by clinical features and then evaluated by laboratory findings (neuronal antibodies, cerebrospinal fluid (CSF), MRI, FDG-PET). The results of preliminary application of the algorithm are described. One of the three neuronal antibodies was positive in 13 patients. In this preliminary investigation, it was suggested that two or more abnormal findings in the diagnostic tests (CSF, MRI, FDG-PET) favors the diagnosis of autoimmune epilepsy. On the other hand, two patients with a positive neuronal antibody test failed the first step (clinical features), partly because epilepsy was not the major manifestation of autoimmune encephalitis (VGKC complex antibody) or due to a relatively low titer of the antibody (GAD antibody). Recruitment of the patient cohort with comprehensive neuronal antibody testing and multivariate analysis of laboratory findings is warranted for validation and modification of the proposed algorithm.