Inoculation of blood culture vials with joint fluid samples has revealed the important pathogenic role of Kingella kingae in pediatric arthritis. However, recent studies based on broad-range 16S ribosomal DNA PCR and real-time PCR without a probe suggest that conventional methods remain suboptimal. We developed a new real-time PCR method with a probe that is highly specific for K. kingae and applied it to joint fluid samples collected from 89 children with suspected arthritis admitted to our institution during a 2-year period. Real-time PCR was also applied to blood samples obtained before surgery and to joint drainage fluid samples obtained during several days after surgery. Thirty-six (40%) of the 89 cases of suspected septic arthritis had positive culture. Staphylococcus aureus was the main isolate (n = 19/36, 53%), followed by K. kingae (n = 7/36, 19%). Specific real-time PCR identified K. kingae in 24 of the 53 culture-negative cases. Thus, K. kingae was present in 31 (52%) of the 60 documented cases, making it the leading pathogen. Real-time PCR on all 15 blood DNA extracts from patients with K. kingae infection was negative, demonstrating that joint fluid positivity did not result from DNA circulating in blood. Real-time PCR amplification of drainage fluid samples showed that the pathogen could be detected for up to 6 days after antibiotic initiation. K. kingae real-time PCR applied to DNA extracted from joint fluid samples, but not from blood samples, markedly improved the etiological diagnosis of septic arthritis in children. Retrospective diagnosis is feasible for up to 6 days after treatment initiation.