Objectives Diagnosis of feline infectious peritonitis (FIP) remains challenging, especially in cats without effusions. The objective of this study was to evaluate the sensitivity and specificity of a real-time reverse transcriptase polymerase chain reaction (RT-PCR) detecting feline coronavirus (FCoV) RNA in peripheral blood mononuclear cells (PBMCs) and serum in comparison with the same real-time RT-PCR in cell-free body cavity effusion. Methods This prospective case-control study included 92 cats. Forty-three cats had a definitive diagnosis of FIP, established either by histopathological examination (n = 28) or by positive immunofluorescence staining of FCoV antigen in macrophages of effusions (n = 11), or by both methods (n = 4). Forty-nine control cats had other diseases but similar clinical signs. Real-time RT-PCR was performed on PBMCs of 37 cats (21 cats with FIP, 16 controls), on serum of 51 cats (26 cats with FIP, 25 controls) and on cell-free body cavity effusion of 69 cats (36 cats with FIP, 33 controls). Sensitivity, specificity, positive and negative predictive value, including 95% confidence intervals (CI), were calculated. Results Real-time RT-PCR of PBMCs, serum and cell-free body cavity effusion showed a specificity of 100% (95% CI 79.4-100% in PBMCs, 86.3-100% in serum, 89.4-100% in cell-free body cavity effusion) and a sensitivity of 28.6% (95% CI 11.3-52.2%) in PBMCs, 15.4% (95% CI 4.4-34.9%) in serum and 88.9% (95% CI 73.9-96.9%) in cell-free body cavity effusion to diagnose FIP. Conclusions and relevance Although it is known that RT-PCR can often provide false-positive results in healthy cats, this real-time RT-PCR was shown to be a specific tool for the diagnosis of FIP when applied in a clinical setting. Sensitivity in cell-free body cavity effusion was high but low in PBMCs and serum. PBMC samples showed a higher sensitivity than serum samples, and are therefore a better choice if no effusion is present.