Background Natural orifice transluminal endoscopic surgery (NOTES) requires fast and steady CO2 insufflation into the intraluminal and intra-abdominal spaces through a flexible endoscope. However, an optimal endoscopic insufflation system has yet to be determined. Objective To verify the performances of 2 currently available CO2 insufflators in an experimental NOTES setting: (1) an automatic pressure-regulated surgical insufflator (UHI-3) and (2) a manual endoscopic insufflator (UCR). Design An inanimate bench study followed by an acute animal experiment. Setting Osaka University and Olympus Research and Development Department. Main Outcome Measurements The UHI-3 or UCR was connected to an endoscope of differing length and diameter via an insufflating line of differing length and diameter. The flow rates at the tip of the endoscope (bench test), the time to establish pneumoperitoneum, and the time to re-establish pneumoperitoneum after forceful suction (porcine model) were obtained. Results The UHI-3 failed to feed CO2 through an insufflating channel but fed CO2 via a working channel but required a large channel (>3 mm) and a wide insufflating line (>7 mm) to accomplish an acceptable flow rate. UCR fed CO2 through the insufflating channel; however, the time taken to establish pneumoperitoneum and the time taken to re-establish pneumoperitoneum after forceful suction were longer compared with the time taken for UHI-3 insufflation via the working channel or laparoscopic cannula. Limitations Bench/animal study with small sample numbers; no human trial. Conclusions The currently available CO2 insufflators are not optimal for NOTES. Modification of an endoscopic insufflation system and/or development of a dedicated overtube with an insufflating function are therefore essential.