[Purpose] Gait-training robots have been developed for stroke patients with gait disturbance. It is important to survey the views of physiatrists and physical therapists on the characteristics of these devices during their development. [Subjects and Methods] A total of 100 physiatrists and 100 physical therapists from 38 hospitals participated in our questionnaire survey. [Results] The most common answers about the merits of gait-training robots concern improving the treatment effects (28.5%), followed by standardizing treatment (19%), motivating patients about treatment (17%), and improving patients’ self-esteem (14%). The subacute period (1–3 months post-stroke onset) was most often chosen as the ideal period (47.3%) for the use of these devices, and a functional ambulation classification of 0–2 was the most selected response for the optimal patient status (27%). The preferred model was the treadmill type (47.5%) over the overground walking type (40%). The most favored commercial price was $50,000–$100,000 (38.3%). The most selected optimal duration for robot-assisted gait therapy was 30–45 min (47%), followed by 15–30 min (29%), 45–60 min (18%), ≥ 60 min (5%), and < 15 min (1%). [Conclusion] Our study findings could guide the future designs of more effective gait-training robots for stroke patients.