The clinical utility of rapid multiplex respiratory specimen PCR panels for pneumonia for patients with suspected pneumonia is undefined. We aimed to compare the effect of the BioFire FilmArraypneumonia panel (bioMérieux, Salt Lake City, UT, USA) with standard of care testing on antibiotic use in a real-world hospital setting. We conducted a single-centre, open-label, pragmatic, randomised controlled trial at the Mayo Clinic, Rochester, MN, USA. Hospitalised patients (aged ≥18years) with suspected pneumonia, from whom expectorated or induced sputum, tracheal secretions, or bronchoalveolar lavage fluid respiratory culture samples (one per individual) could be collected during index hospitalisation, were eligible for inclusion. Samples from eligible participants were randomly assigned (1:1) with a computerised tool to undergo testing with either the BioFire FilmArray pneumonia panel, conventional culture, and antimicrobial susceptibility testing (intervention group) or conventional culture and antimicrobial susceptibility testing alone (control group). Antimicrobial stewardship review in both groups involved an assessment and recommendations for antibiotic modifications based on clinical data and the results from the BioFire FilmArray pneumonia panel, conventional culture, or both. The primary outcome was median time to first antibiotic modification (ie, escalation or de-escalation of antibiotics against Gram-negative and Gram-positive bacteria) within 96h of randomisation, assessed with the Wilcoxon rank-sum test and analysed in a modified intention-to-treat population. This trial is registered with ClinicalTrials.gov (NCT05937126). Between Sept 15, 2020, and Sept 19, 2022, 1547patients were screened for eligibility, of whom 1181 (76·3%) were randomly assigned: 582 (49·3%) to the intervention group and 599 (50·7%) to the control group. In total, 1152participants were included in the modified intention-to-treat analysis, 589 (51·1%) in the control group and 563(48·9%) in the intervention group. For the modified intention-to-treat population, median time to any first antibiotic modification was 20·4h (95% CI 18·0-20·4) in the intervention group and 25·8h (22·0-28·7) in the control group (p=0·076). Median time to any antibiotic escalation was 13·8h (9·2-19·0) in the intervention group and 24·1h (19·5-29·6) in the control group (p=0·0022). Median time to escalation of antibiotics against Gram-positive organisms was 10·3h (6·2-30·9) in the intervention group and 24·6h (19·5-37·2) in the control group (p=0·044); median time to escalation of antibiotics against Gram-negative organisms was 17·3h (10·8-23·3) in the intervention group and 27·2h (21·3-33·9) in the control group (p=0·010). Median time to any antibiotic de-escalation did not differ between groups (p=0·37). Median time to first de-escalation of antibiotics against Gram-positive organisms was 20·7h (17·8-24·0) in the intervention group and 27·8h (22·9-33·0) in the control group (p=0·015); median time to first de-escalation of antibiotics against Gram-negative organisms did not differ between groups (p=0·46). Clinical use of the BioFire FilmArray pneumonia panel might lead to faster antibiotic escalations, including for Gram-negative or Gram-positive bacteria, and faster antibiotic de-escalations directed at Gram-positive bacteria. Additional research is needed regarding antimicrobial de-escalation, especially when antibiotics with broad Gram-negative spectrum are being used, by use of rapid diagnostics in patients with lower respiratory tract infection. bioMérieux.