•This randomised controlled trial compared robot-assisted gait training (RAGT) to treadmill training (TT) for people with Parkinson's disease.•RAGT and TT improved overall gait capacity to the same extent.•RAGT was more effective than TT in improving freezing of gait (FOG).•RAGT reduced FOG more in dependent walkers than independent walkers.•Disability level was reduced in freezers more than in non-freezers. Although gait disorders strongly contribute to perceived disability in people with Parkinson's disease, clinical trials have failed to identify which task-oriented gait training method can provide the best benefit. Freezing of gait remains one of the least investigated and most troublesome symptoms. We aimed to compare the effects of robot-assisted gait training and treadmill training on endurance and gait capacity in people with Parkinson disease; the secondary aim was to compare the effect of the treatments in people with freezing and/or severe gait disability and assess changes in overall disease-related disability and quality of life. Outpatients with Parkinson disease (Hoehn and Yahr stage≥2) were randomly assigned to receive 20 sessions of 45-min gait training assisted by an end-effector robotic device (G-EO System) or treadmill training. Outcome assessments were the 6-min walk test, Timed Up and Go test, Freezing of Gait Questionnaire, Unified Parkinson's Disease Rating Scales and Parkinson's Disease Quality of Life Questionnaire-39 administered before (T0) and after treatment (T1). We included 96 individuals with Parkinson disease: 48 with robot-assisted gait training and 48 treadmill training. Both groups showed significant improvement in all outcomes. As compared with baseline, with robot-assisted gait training and treadmill training, endurance and gait capacity were enhanced by 18% and 12%, respectively, and motor symptoms and quality of life were improved by 17% and 15%. The maximum advantage was observed with the Freezing of Gait Questionnaire score, which decreased by 20% after either treatment. On post-hoc analysis, dependent walkers benefited more than independent walkers from any gait training, whereas freezers gained more from robot-assisted than treadmill training in terms of freezing reduction. Repetitive intensive gait training is an effective treatment for people with Parkinson disease and can increase endurance and gait velocity, especially for those with severe walking disability. Advantages are greater with robot-assisted gait training than treadmill training for individuals with freezing of gait – related disability.