Purpose: This systematic review aims to determine whether robot-assisted training is more effective in gait training for persons with subacute hemiparetic stroke. Methods: This study adopted a systematic review study design focused on subacute hemiparetic stroke, and four core academic databases were searched until June 11, 2021, for relevant studies, including PubMed, Embase, the Cochrane Library, and ProQuest Central. The review included randomized controlled trials (RCTs) evaluating the effects of robotic-assisted training on gait performance in persons with a diagnosis of subacute hemiparetic stroke. The selected RCT studies were qualitatively synthesized based on the population, intervention, comparison, outcome, settings, and study design (PICOS-SD). Results: The study selected five RCTs involving 253 subacute hemiparetic stroke patients and performing robotic-assisted gait training using the following devices: the Lokomat, Morning Walk, Walkbot, ProStep Plus, or Gait Trainer II. Five RCTs were eligible for the meta-analysis after quantitative synthesis, and the results showed that the robot-assisted gait training group had a greater gait performance than the control group based on the 10-meter walk test, Berg balance scale, Rivermed mobility index, functional ambulation category, and modified Barthel index. Conclusion: The results of this study showed that the gait performance of subacute hemiparetic stroke patients changes throughout robot-assisted gait training, but there were no indications that any of the clinically relevant effects of robot-assisted training are greater than those of conventional gait training. Further, the small sample size and different therapeutic intensities indicate that definitive conclusions could not be made.