Background Periprosthetic joint infection (PJI) remains challenging since a “gold standard” for diagnosis has not yet been established. This study aimed to evaluate the accuracy of synovial fluid procalcitonin (SF-PCT) and serum procalcitonin as a diagnostic biomarker for PJI and to compare its accuracy against standard methods. Methods A prospective cohort study was conducted during 2015–2017 in 32 patients with painful hip or knee arthroplasty who have underwent revision surgery. Relevant clinical and laboratory data were collected. PJI was diagnosed based on the 2013 international consensus criteria. Preoperative blood sample and intraoperatively acquired joint fluid were taken for PCT measurement with a standard assay. Diagnostic accuracy was analyzed by the receiver-operating characteristic curve and the area under the curve (AUC). Results Twenty patients (62.5%) were classified as the PJI group, and 12 (37.5%) were classified as the aseptic loosening group. The median age was 68 years (range 38–87 years). The median values of SF-PCT and serum PCT in the PJI group were both significantly higher than those in the aseptic loosening group: the median serum PCT levels (interquartile range: IQR) were 0.33 ng/mL (0.08-2.79 ng/mL) in the PJI group compared with 0.04 ng/mL (0.03-0.06 ng/mL), and the median SF-PCT levels (IQR) were 0.16 ng/mL (0.12-0.26 ng/mL) in PJI group compared with 0.00 (0.00-0.00 ng/mL) (p < 0.001, both). SF-PCT, with a cut-off level of 0.08 ng/mL, had an AUC of 0.87, a sensitivity of 90.0%, a specificity of 83.3%, and a negative likelihood ratio (LR-) of 0.12. Serum PCT, with a standard cut-off level of 0.5 ng/mL, had an AUC of 0.70, a sensitivity of 40.0%, a specificity of 100.0%, and a LR- of 0.60. Conclusion SF-PCT appears to be a reliable test and could be useful as an alternative indicator or in combination with standard methods for diagnosing PJI.