Objective To compare the overall efficacy of procalcitonin （PCT）-guided antimicrobial therapy with that of conventional antimicrobial therapy in patients with sepsis， and to analyze the interaction and differential benefits of the PCT-guided strategy among subgroups with different clinical characteristics， thereby providing real-world evidence for individualized clinical application. Methods A total of 515 patients with sepsis who were admitted to Sun Yat-sen Memorial Hospital of Sun Yat-sen University from January 2021 to June 2024 were included in this retrospective cohort study. According to the actual monitoring and treatment strategies implemented in clinical practice， the patients were divided into a PCT-guided group （n = 245） and a conventional treatment group （n = 270）. The baseline characteristics， such as age， sex， and underlying diseases， were well-balanced between the two groups. In the PCT-guided group， the PCT levels were dynamically monitored every 24 - 48 hours for at least three times. The core antibiotic discontinuation threshold was defined as a ≥ 80% decline in the peak PCT level or two consecutive PCT values < 0.5 ng/mL， and the antimicrobial therapy was adjusted accordingly in combination with the clinical assessment. In contrast， the conventional treatment group received antibiotic adjustments based on traditional indicators and clinical experience. Subsequently， the overall core efficacy and safety outcomes were compared between the two groups. Subgroup analyses were carried out based on age， Sequential Organ Failure Assessment （SOFA） score， and the presence of diabetes mellitus. A regression model-based interaction term test was employed to examine the interaction effects of each clinical characteristic with the PCT-guided strategy. Results The PCT-guided group exhibited a significantly shorter duration of antibiotic use ［408.01 （228.06， 887.00） h vs. 579.01 （288.99， 1 290.77） h， P = 0.001］ when compared with the conventional group. The PCT level at the time of antibiotic discontinuation ［0.31 （0.09， 1.41） ng/mL vs. 0.97 （0.22， 6.04） ng/mL， P < 0.001］， antibiotic costs ［6 790.44 （2 393.00， 21 712.71） yuan vs. 12 231.96 （3 396.04， 27 211.60） yuan， P = 0.004］， and the 28-day mortality rate （16.7% vs. 28.1%， P = 0.003） were also notably lower in the PCT-guided group. No significant differences were detected between the two groups in terms of the average daily antibiotic cost or the incidence of kidney dysfunction （all P > 0.05）. Subgroup and interaction analyses indicated that there was a significant interaction between age and the PCT-guided strategy regarding antibiotic duration （P for interaction = 0.039）， and between diabetes status and the PCT-guided strategy regarding the length of hospital stay （P for interaction = 0.003）； no significant interaction was identified between the SOFA score and the PCT-guided strategy （all P for interaction > 0.05）. Conclusions Dynamic monitoring of PCT can assist in shortening the duration of antibiotic therapy， reducing medical costs， and enhancing short-term prognosis. This benefit is more pronounced in patients aged less than 65 years and without diabetes. Elderly patients and those with diabetes should undergo comprehensive clinical evaluation.