Hand exoskeleton-type precision grip is a form of robotic development that attempts to imitate human finger movement to rehabilitate post-stroke patients. The exoskeletons can manipulate individual fingers through precise movements; nevertheless, their mechanical structure is intricate, resulting in reduced practicality and increased weight during usage. One way to fix this is to make the exoskeleton lighter so the finger can move quickly and have a small load. This research aims to reduce the mass of the hand exoskeleton by optimizing the thumb's structure by applying generative design methods. The generative design optimization process produces 60 outcomes, evaluated and ranked according to their design solution status, recommendation values, visual aesthetics, and mechanical performance validation. The best design is the O-1 design, which has a software mass reduction of 74% to 76%, a mass reduction after manufacturing of 67%, the highest safety factor of 1,725, a minimum stress of 11.59 MPa, and a displacement of 0.7288 mm. The O-1 design is the best one that the generative design method came up with compared to the other designs.