Most patients diagnosed with oral squamous cell carcinoma (OSCC) present with locally advanced stages, which are typically associated with poor outcomes. Although immunotherapy offers potential improvements in patient survival, its efficacy is hampered by low response rates. The microbiome is widely involved in tumor immunity and may play a role in immunotherapy. This study aimed to investigate the potential association between the oral (salivary) microbiome and immunotherapy response in patients with OSCC. Salivary metagenome sequencing was performed on 47 patients with OSCC undergoing neoadjuvant immunotherapy (NAIT) in a clinical trial (NCT04649476). Patients were divided into responders and nonresponders based on their pathological responses. The results showed that the species richness of the salivary microbiome was lower in the nonresponders before NAIT than in the responders. Differential analysis revealed that nonresponders exhibited a lower relative abundance of 34 bacterial species and a higher relative abundance of 4 bacterial species. Notably, low levels of Eubacterium infirmum, Actinobaculum, and Selenomas (EAS) in the saliva may be associated with the nonresponse of patients with OSCC to NAIT. A nomogram based on EAS was developed and validated to determine the efficacy of NAIT. The area under the curve for the training cohort was 0.81 (95% confidence interval, 0.66 to 0.81). Quantitative polymerase chain reaction confirmed that low levels of salivary EAS effectively identified nonresponders to NAIT. Furthermore, the low abundance of salivary EAS was closely correlated with a low density of intratumoral CD4+, CD14+, CD68+, and FOXP3+ cells. Metabolic functional annotation revealed numerous biosynthetic processes associated with EAS that were more active in responders. In summary, this study provides valuable data resources for the salivary microbiome and reveals that nonresponders have different salivary microbiome profiles than responders do before NAIT. Low salivary EAS levels can serve as potential biomarkers for distinguishing nonresponders from responders.