Neonatal sepsis is a major cause of worldwide morbidity and mortality. Blood cultures are considered the gold standard for diagnosis, but results are often delayed for 24 to 48 hours, and sensitivity, although improved by modern techniques, such as automated blood cultures, is variable and affected by the bacterial load. For these reasons, empiric antibiotics are frequently administered to avoid potential devastating consequences of untreated sepsis. Unnecessary antibiotic treatment has been associated with increased mortality and other adverse outcomes; therefore, antibiotics should be discontinued as soon as sepsis has been ruled out. Negative cultures pose a challenge to clinicians, who must distinguish between real sepsis and sepsis-like conditions (noninfectious or viral) which do not require antibiotics. Focal infections with negative blood cultures do require antibiotic treatment. Ultra-low bacteremia, primary or secondary to recent antibiotic exposure, is often associated with negative cultures, and some consider a short course of empiric antibiotics sufficient for clearing of bacteremia. Biomarkers and molecular methods based on polymerase chain reaction are important add-ons to clinical signs or symptoms for establishing the diagnosis of sepsis. Other promising future potential adjuvants are metabolomics. Antibiotic stewardship should be implemented to avoid or discontinue unnecessary treatment. Prevention of infection still remains the most important step for dealing with neonatal sepsis. KEY POINTS: · Blood cultures are the gold standard diagnosis of neonatal sepsis but sometimes may be negative.. · Other bacterial, viral, and noninfectious conditions may mimic sepsis, prompting initiation of empiric antibiotic treatments.. · Since a definition of neonatal sepsis is lacking, recognizing real septic episodes may be challenging.