Background Robot‐assisted gastrectomy (RAG) has been developed in the hope of improving surgical quality and overcoming the limitations of conventional laparoscopically assisted gastrectomy (LAG) and open gastrectomy (OG) for gastric cancer. The aim of this study was to determine the extent of evidence in support of these ideals. Methods A systematic review of the three operation types (RAG, LAG and OG) was carried out to evaluate short‐term outcomes including duration of operation, retrieved lymph nodes, estimated blood loss, resection margin status, technical postoperative complications and hospital stay. Results Nine non‐randomized observational clinical studies involving 7200 patients satisfied the eligibility criteria. RAG was associated with longer operating times than LAG and OG (weighted mean difference 61·99 and 65·73 min respectively; P ≤ 0·001). The number of retrieved lymph nodes and the resection margin length in RAG were comparable with those of LAG and OG. Estimated blood loss was significantly less in RAG than in OG (P = 0·002), but not LAG. Mean hospital stay for RAG was similar to that for LAG (P = 0·14). In contrast, hospital stay was significantly shorter, by a mean of 2·18 days, for RAG compared with OG (P < 0·001). Postoperative complications were similar for all three operative approaches. Conclusion Short‐term oncological outcomes of RAG were comparable with those of the other approaches. LAG was a shorter procedure and less expensive than RAG. Future studies involving RAG should focus on minimizing duration of operation and reducing cost.