We developed a novel training simulator for percutaneous endoscopic gastrostomy (PEG) and determined its efficacy and realism in PEG insertion training. The gastrostomy simulator was made using three-dimensional printing and silicone molding technology. The simulator is of two types: pull type and introducer type. We enrolled 20 participants, comprising 10 beginners with no experience of PEG insertion and 10 PEG-experienced endoscopists. Each participant underwent two training sessions for the two simulator types. We recorded the simulation time, self-evaluation, and difficulty score based on a 5-score scale for PEG insertion among the participants. Subsequently, simulator performance was assessed via a questionnaire based on a 7-point Likert scale. The mean time to completion of PEG simulation decreased from 11.9 (5.2) to 9.0 (4.0) min for the pull type and from 13.8 (7.0) to 12.0 (5.8) min for the introducer type in the beginner group. The mean self-evaluation scores of beginners increased from 2.2 (1.1) to 3.1 (0.7) (pull type) and from 2.2 (1.2) to 3.3 (0.8) (introducer type). The mean procedure difficulty scores of beginners decreased from 3.4 (1.1) to 2.7 (0.9) (pull type) and from 4.4 (0.5) to 3.0 (0.8) (introducer type). The improvement of skill score was 6.3 (1.2) for the beginner group. The general realism score of the simulator for handling was judged to be 6.0 (0.9) by the experienced group. The three-dimensional-printed simulator for PEG insertion can be useful for training of beginner endoscopists and shows good efficacy and realism.