Background: The differential diagnosis of autoimmune and infectious encephalitis is notoriously difficult. For this study, we compare the presenting clinical symptoms and paraclinical test results of autoimmune and infectious encephalitis patients. A clinical algorithm for the diagnosis of autoimmune encephalitis has recently been published. We test these Graus criteria on our cohort for diagnostic sensitivity and specificity within the first week of presentation.Methods: We included all patients seen at our department within a 10-year-period who were diagnosed with encephalitis. The discharge diagnoses served as the reference standard for testing the clinical algorithm for two conditions: use of all the clinical information available on a patient during the first week of hospital admission assuming undefined autoantibody status and microbiological test results (C1) vs. consideration of all the information available on a patient, including the results of serological and microbiological testing (C2).Results: Eighty-four patients (33 autoimmune, 51 infectious encephalitis) were included in the study. Fifty-one (17 autoimmune, 34 infectious) had a definite clinical diagnosis. The two groups differed significantly for the presence of headache, fever, epileptic seizures, and CSF cell-count at presentation. Application of the clinical algorithm resulted in a low sensitivity (58%) and very low specificity (8%) for the diagnosis of possible autoimmune encephalitis. The latter increased considerably in the subgroups of probable and definite autoimmune encephalitis. Whereas the sensitivity of the individual diagnostic categories was clearly time-dependent, the specificity rested foremost on the knowledge of the results of microbiological testing. Anti-CASPR2- and -LGI1-associated autoimmune encephalitis and tick-borne virus encephalitis presented particular diagnostic pitfalls.Conclusions: We define clinical symptoms and paraclinical test results which prove valuable for the differentiation between infectious and autoimmune encephalitis. Sensitivity and specificity of the clinical algorithm clearly depended on the amount of time passed after hospital admission and knowledge of microbiological test results. Accepting this limitation for the acute setting, the algorithm remains a valuable diagnostic aid for antibody-negative autoimmune encephalitis or in resource-poor settings. The initiation of immune therapy however should not be delayed if an autoimmune etiology is considered likely, even if the diagnostic criteria of the algorithm are not (yet) fulfilled.