AbstractAimWe compared three different intensities of robot‐assisted gait training (RAGT) for achieving favourable outcomes in children with cerebral palsy (CP).MethodThis study was conducted using a randomized controlled, single‐blind design. Thirty children (19 males and 11 females; mean age 6 years 1 month, SD 2 years) with CP classified in Gross Motor Function Classification System levels II and III were assigned to three different RAGT intensity groups: high‐intensity (fastest walking speed and lowest body weight support [BWS]), low‐intensity (slowest speed and highest BWS), and comfortable intensity (intermediate speed and intermediate BWS). The RAGT intervention was performed three times a week for 6 weeks. Outcome measures included the 88‐item Gross Motor Function Measure, stability index, spatiotemporal parameters of gait analysis, paediatric functional independence measure, and the Canadian Occupational Performance Measure.ResultsThe 88‐item Gross Motor Function Measure was significantly improved after training in the high‐intensity (D Δ8.3 ± 15.6; E Δ3.8 ± 4.1) and comfortable intensity (D Δ2.9 ± 3.1; E Δ1.2 ± 2.0) groups, whereas gait speed was improved in the comfortable intensity group, without statistically significant group differences. Only the low‐intensity group showed improvement on the stability index (Δ −0.6 ± 0.9, p = 0.05). Everyday functional performance significantly improved in all three groups, with the comfortable intensity group showing the greatest improvement.InterpretationDifferent training intensities produced improvement in different areas; individualized RAGT intensity adjustment is therefore needed based on the rehabilitation goal.