We prospectively evaluated the diagnostic accuracy of cerebrospinal fluid (CSF)-beta-amyloid1-42 (Abeta42), -total-tau (tau) and -phosphorylated-tau181 (p-tau181) as measured by sandwich ELISAs in the clinical routine of a community state hospital to discriminate between patients with Alzheimer's disease (AD), healthy controls (HC), non-AD-dementias, a group composed of various psychiatric disorders (non-AD-dementias, mental diseases) and an age-matched random sample (RS) (total N=219). By comparing patients with AD to HC as reference, tau revealed sensitivity (sens)/specificity (spec) of 88%/80%, p-tau(181) 88%/80%, tau/Abeta42-ratio 81%/85% and phospho-tau(181)/Abeta42-ratio 81%/78%. Discriminative power between HC and all dementias under investigation was estimated lower for tau (78%/77%) and p-tau(181) (73%/79%). Relative to patients with AD, ROC analysis for the RS revealed highest sens/spec for p-tau181 (79%/77%) and p-tau181/Abeta42 ratio (78%/75%). Differentiation between AD versus a group made of patients with various psychiatric disorders was optimised by using CSF-p-tau181 (80%/77%). Under clinical routine conditions current CSF-biomarkers show a substantial capacity to discriminate between AD and HC as reference and to mark off AD patients from RS and heterogeneous diagnostic groups composed of non-AD dementias and other psychiatric conditions. Despite a residual substantial overlap between the groups, we conclude that current CSF markers are well suited to support AD-related diagnostic procedures in every-day clinics.