The study aims to evaluate the significance of ARID1A expression in gastric cancer (GC) patients receiving neoadjuvant therapy, using immunohistochemical methods and bioinformatics approaches. Immunohistochemical analysis of an institutional clinical cohort (n = 80) was combined with bioinformatics analyses of independent public datasets. RNA-sequencing and clinical data from the TCGA-STAD cohort were retrieved and processed using TCGAbiolinks, differential expression analysis was performed using DESeq2, and overall survival (OS) was evaluated using Kaplan-Meier, log-rank, and univariate Cox regression analyses. STRING was used to construct an ARID1A-centered protein association network, while Human Protein Atlas and Cellosaurus data were used to evaluate ARID1A expression. Loss of ARID1A protein expression rate was 41.3%. ARID1A loss and larger tumor size were correlated (median 6.0 cm vs. 4.0 cm; p = 0.004). ARID1A status and histopathological response to neoadjuvant therapy, pathological tumor stage and OS are not statistically related. Bioinformatics analyses demonstrated marked variation in ARID1A expression across GC cell lines, while TCGA-STAD analyses showed no significant tumor-normal difference or survival association at the transcript level. While ARID1A loss is associated with increased tumor size, it is not a strong independent predictor of treatment response or OS in GC. These findings underscore the molecular complexity of gastric adenocarcinoma and the need for larger, multicenter studies to refine the clinical utility of ARID1A.