ObjectiveMetagenomic next-generation sequencing (mNGS) is emerging to be pivotal in infectious disease diagnosis, particularly in pulmonary infection. However, the utility of bronchoalveolar lavage fluid (BALF) versus blood mNGS remains controversial. Therefore, the current meta-analysis summarized the previous studies regarding BALF or blood mNGS in pulmonary infection, aiming to comprehensively compare the diagnostic efficiency between them.MethodsStudies reporting paired BALF and blood mNGS data for pulmonary infection diagnosis were searched in the PubMed, Web of Science, Embase, CNKI, and Wanfang from January 2016 to March 2024.ResultsEleven studies involving 346 assessed patients were eligible and analyzed. The pooled sensitivity, specificity, and area under the curve (AUC) of BALF mNGS were 0.94 [95% confidence interval (CI): 0.90–0.97], 0.27 (95% CI: 0.19–0.36), and 0.86 [standard error (SE): 0.06], respectively. Meanwhile, the pooled sensitivity, specificity, and AUC of blood mNGS were 0.64 (95% CI: 0.56–0.72), 0.69 (95% CI: 0.62–0.76), and 0.81 (SE: 0.05), respectively. By subgroup analyses, the AUCs of BALF mNGS and blood mNGS for viral detection were 0.70 (SE: 0.08) and 0.71 (SE: 0.08), respectively, while the AUCs of BALF mNGS and blood mNGS for nonviral (bacterial or fungal) detection were 0.83 (SE: 0.06) and 0.73 (SE: 0.08), respectively. Moreover, no threshold effect or publication bias existed, and sensitivity analysis revealed that the findings were generally robust.ConclusionBALF mNGS surpasses blood mNGS for total pathogen detection in pulmonary infection patients, while they share a similar efficiency for viral detection.Systematic review registrationhttps://www.crd.york.ac.uk/PROSPERO/, identifier CRD42024562740.