<p indent="0mm">Autoimmune liver disease (AILD) is a kind of liver disease caused by immune dysfunction, including autoimmune hepatitis (AIH), primary biliary cholangitis (PBC), and overlap syndrome (OS). In AILD patients, the immune system attacks the liver, and elevated autoantibodies can be found in serum. AIH and PBC are two kinds of AILD with high incidence, and their pathogenesis is different. AIH is characterized by interfacial hepatitis caused by the autoimmune system attacking hepatocytes. The clinical features of PBC are non-suppurative, destructive cholestasis resulting from immune damage affecting the interlobular bile ducts. Due to the complex and promiscuous pathogenesis of autoimmune diseases and the low specificity of clinical features, challenges remain in the clinical diagnosis of AILD. Liver biopsy results are necessary for the accurate diagnosis of AILD, but biopsy is not conducive to the prognosis of patients, and the pathological changes of patients may be non-specific. Therefore, the need of finding serological metabolites with low high specificity and invasive features becomes urgent. Metabolomics has unique advantages in reflecting the overall pathophysiological changes of the body by analyzing the metabolites. Untargeted metabolomics has been widely used in the study of liver-related diseases and has good application value for clinical diagnosis. In this study, we attempted to characterize the metabolic signatures of autoimmune liver diseases by serum metabolomics. A mass spectrometry-based untargeted metabolomics study was performed on the serum of AIH (<italic>n</italic>=42), PBC (<italic>n</italic>=103), and AIH-PBC OS (<italic>n</italic>=57) patients. The differentiable metabolites for PBC and AIH samples were obtained by principal component analysis and orthogonal partial least-squares-discriminant analysis, which consisted of the metabolomic signatures. Among the metabolomic signatures, the fingerprint metabolites were further extracted by clustering analysis using the values of area under the curve (AUC) of each differentiable metabolite, binary logistic regression model, and variable importance projection scores. 101 signature metabolites from all metabolite variables were identified, and 29 fingerprint metabolites were further screened out and enriched in bile, phospholipids, and amino acids metabolisms pathway, with most classification ability to distinguish PBC and AIH. The eigenmetabolite compressed from the fingerprint metabolites visualized the significant difference between PBC and AIH (<italic>P</italic><0.0001) and achieved a good potential for classification (AUC=0.797). Based on fingerprint metabolites, AIH and PBC samples could be separated into three zones, the “pure zone” of PBC, the “pure zone” of AIH, and the mixture zone of PBC and AIH. We put OS samples into the model and found that samples showed different distribution characteristics in three zones, indicating that OS patients according to the current diagnostic criteria have different metabolic characteristics and etiological tendencies, requiring a more accurate diagnosis. This study revealed a metabolic outlook for AIH, PBC, and their overlap syndrome, and the diagnostic model based on fingerprint metabolites with the potential ability to facilitate precise diagnosis and management of patients with autoimmune liver diseases.