Epstein-Barr virus (EBV)-associated gastric cancer (EBVaGC), whose prognosis remains controversial, is diagnosed by in situ hybridization of EBV-derived EBER1/2 small RNAs. In the Cancer Genome Atlas (TCGA)-Stomach Adenocarcinoma (STAD) project, the EBV molecular subtype was determined through a combination of multiple next-generation sequencing methods, but not by the gold standard in situ hybridization method. This leaves unanswered questions regarding the discordance of EBV positivity detected by different approaches and the threshold of sequencing reads. Therefore, we re-analyzed the TCGA-STAD RNA-seq dataset including 375 tumor and 32 normal samples, using our analysis pipeline. We defined a reliable threshold for EBV-derived next-generation sequencing reads by mapping them to the EBV genome with three different random arbitrary alignments. We analyzed the prognostic impact of EBV status on the histopathological subtypes of gastric cancer. EBV-positive cases identified by re-analysis comprised nearly half of the cases (49.6%) independent from infiltrating lymphocyte signatures, and showed significantly longer overall survival for adenocarcinomas of the "not otherwise specified" (NOS) type [P = 0.016 (log-rank test); HR 0.476; 95% CI 0.260-0.870, P = 0.016 (Cox univariate analysis)], but shorter overall survival for the tubular adenocarcinoma type [P = 0.005 (log-rank test); HR 3.329; 95% CI 1.406-7.885, P = 0.006 (Cox univariate analysis)]. These results demonstrate that the EBV-positivity rates were higher when determined by RNA-seq than when determined by EBER1/2 in situ hybridization. The RNA-seq-based EBV-positivity demonstrated distinct results for gastric cancer prognosis depending on the histopathological subtype, suggesting its potential to be used in clinical prognoses.