Purpose The present study aimed to establish pretreatment protocols as well as real-time and droplet digital polymerase chain reaction (PCR) methodologies to detect and quantify Erysipelothrix rhusiopathiae (ER) DNA in blood samples from infected chickens, as tools for routine diagnostics and monitoring of experimental infections. Chicken blood is a problematic matrix for PCR analysis because nucleated erythrocytes contribute large amounts of host DNA that inhibit amplification. Methodology Using artificially spiked samples of fresh chicken blood, as well as blood samples from three experimental infection studies, the performance of pretreatment protocols, including choice of blood stabilization agent, centrifugation speeds and Ficoll gradient separation, was evaluated. The results were compared with those from traditional culture-based protocols combined with matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS). Results/Key findings Simple preparations producing cell-free samples performed well on artificial spike-in samples, providing high sensitivity. However, performance was poor in clinical samples or artificial samples where the bacteria were incubated for 4 h or more in fresh blood prior to DNA extraction. In these samples, a Ficoll separation protocol that creates samples rich in lymphocytes, monocytes and thrombocytes prior to DNA extraction was far more effective. Conclusions Our results indicate that ER bacteria undergo rapid phagocytosis in chicken blood and that analysis of a blood fraction enriched for phagocytic cells is necessary for reliable detection and quantification. The presented results explain the poor performance of PCR detection reported in previously published experimental ER infection studies, and the proposed solutions are likely to have broader implications for PCR-based veterinary diagnostics in non-mammalian host species such as poultry and fish.