Objective To explore the causal relationship between social stress and tinnitus onset using a two-sample Mendelian randomization approach. Methods Genetic data pertaining to social stress and tinnitus were extracted from genome-wide association studies (GWAS) databases. Single nucleotide polymorphisms (SNP) that were independent and strongly correlated with social stress were selected as instrumental variables, with tinnitus serving as the outcome variable. Mendelian randomization analyses were performed using inverse-variance weighted (IVW) method, weighted median method, and Mendelian randomization-Egger regression. The intercept term from Mendelian randomization-Egger regression was utilized to assess horizontal pleiotropy, Cochran's Q statistic was employed to evaluate heterogeneity, and leave-one-out analysis was conducted for sensitivity assessment. Results The social stress dataset encompassed 459, 742 samples, while the tinnitus dataset comprised 117, 882 samples. A total of 10 SNPs tightly associated with social stress were identified as instrumental variables. The analysis results of random-effects inverse variance weighting (IVW), weighted median method, and Mendelian randomization-Egger regression were similar (OR=1.251, 1.274, 1.438), indicating that social stress was a risk factor for tinnitus, and there was a positive causal effect between them, with no heterogeneity or pleiotropy in the results. Conclusion There is a causal relationship between social stress and tinnitus onset, which may offer novel perspectives for the clinical prevention and treatment of tinnitus in the future.