QUESTIONS: Does mechanically assisted walking improve walking speed, distance and participation compared with no/non-walking intervention or overground walking after stroke? Are any benefits maintained beyond the intervention period? DESIGN: Systematic review of randomised trials with meta-analysis. PARTICIPANTS: Ambulatory adults at any time after stroke. INTERVENTION: Mechanically assisted walking (treadmill or gait trainer) without body weight support. OUTCOME MEASURES: Walking speed, walking distance and participation. RESULTS: Sixteen trials involving 713 participants were included. The mean PEDro score of the trials was 6.3 (range 4 to 8). Treadmill walking increased walking speed by 0.13 m/s (95% CI 0.08 to 0.19) and distance by 46 m (95% CI 24 to 68) compared with no/non-walking intervention; these effects were largely maintained beyond the intervention. Treadmill walking had a similar or better effect on walking speed (MD 0.07 m/s, 95% CI 0.00 to 0.13) and distance (MD 18 m, 95% CI 1 to 36) compared with overground walking. The estimate of the relative effect of treadmill walking compared with overground walking on participation was very imprecise (SMD 0.16, 95% CI -0.15 to 0.48). CONCLUSION: This systematic review provides moderate-quality evidence that the effect of treadmill walking is the same as or better than the effect of overground walking for improving walking speed and distance in ambulatory people after stroke. Long-term effects and carryover benefits to participation remain uncertain. REVIEW REGISTRATION: PROSPERO (CRD42020162778).