Alpha-1-acid glycoprotein (AGP) increases in the blood of cats with feline infectious peritonitis (FIP), a lethal disease caused by feline coronavirus (FCoV). However, the diagnostic potential of AGP might be limited because AGP also increases in pathophysiological conditions other than FIP. In this retrospective study, the diagnostic potential of serum AGP concentration was evaluated on the basis of the pretest probability of disease, according to the Bayesian approach. Serum AGP levels from cats with FIP (group 1; n = 58) and without FIP (group 2; n = 104) were evaluated. Non-FIP cats were further subgrouped as follows: 2a) inflammation (n = 26), 2b) asymptomatic FCoV infection (n = 49), 2c) injection-site sarcoma (n = 19), 2d) postvaccination (n = 7), and 2e) specific pathogen free (n = 3). Standard descriptive analyses by group and empirical receiver-operating characteristic (ROC) curve estimation were performed. Ordinary logistic regression analysis was performed to derive an estimate of the continuous likelihood ratio to produce the posttest probability of disease for any combination of pretest probability and serum AGP value. The comparison of serum AGP levels in the different groups and the analysis of the ROC curve confirmed that serum AGP is a powerful discriminating marker for FIP. The Bayesian approach demonstrated that when the pretest probability of FIP is high, based on history and clinical signs (groups 1 or 2a), moderate serum AGP levels (1.5-2 mg/ml) can discriminate cats with FIP from others, while only high serum AGP levels (>3 mg/ml) can support a diagnosis of FIP in cats with a low pretest probability of disease (groups 2b to 2e).