The role of procalcitonin (PCT), a widely used sepsis biomarker, in critically ill patients with sepsis is undetermined. To investigate the effect of a low PCT cut-off on antibiotic prescription and to describe the relationships between PCT plasma concentration and sepsis severity and mortality. This was a multicenter (11 Australian intensive care units ICUs), prospective, single-blind, randomized controlled trial involving 400 patients with suspected bacterial infection/sepsis and expected to receive antibiotics and stay in ICU longer than 24 hours. The primary outcome was the cumulative number of antibiotics treatment days at Day 28. PCT was measured daily while in the ICU. A PCT algorithm, including 0.1 ng/ml cut-off, determined antibiotic cessation. Published guidelines and antimicrobial stewardship were used in all patients. Primary analysis included 196 (PCT) versus 198 standard care patients. Ninety-three patients in each group had septic shock. The overall median (interquartile range) number of antibiotic treatment days were 9 (6-21) versus 11 (6-22), P = 0.58; in patients with positive pulmonary culture, 11 (7-27) versus 15 (8-27), P = 0.33; and in patients with septic shock, 9 (6-22) versus 11 (6-24), P = 0.64; with an overall 90-day all-cause mortality of 35 (18%) versus 31 (16%), P = 0.54 in the PCT versus standard care, respectively. Using logistic regression, adjusted for age, ventilation status, and positive culture, the decline rate in log(PCT) over the first 72 hours independently predicted hospital and 90-day mortality (odds ratio 95% confidence interval, 2.76 1.10-6.96, P = 0.03; 3.20 1.30-7.89, P = 0.01, respectively). In critically ill adults with undifferentiated infections, a PCT algorithm including 0.1 ng/ml cut-off did not achieve 25% reduction in duration of antibiotic treatment. Clinical trial registered with http://www.anzctr.org.au (ACTRN12610000809033).