Question: Does mechanically assisted walking increase walking speed or distance in ambulatory people with stroke compared with no intervention/non-walking intervention, or with overground walking? Design: Systematic review with meta-analysis of randomised trials. Participants: Ambulatory adults with stroke. Intervention: Mechanically assisted walking (treadmill or gait trainer) without body weight support. Outcome measures: Walking speed measured in m/s during the 10-m Walk Test and walking distance measured in m during the 6-min Walk Test. Results: Nine studies of treadmill training comprising 977 participants were included. Treadmill training resulted in faster walking than no intervention/non-walking intervention immediately after the intervention period (MD 0.14. m/s, 95% CI 0.09 to 0.19) and this was maintained beyond the intervention period (MD -0.12. m/s, 95% CI 0.08 to 0.17). It also resulted in greater walking distance immediately after the intervention period (MD 40. m, 95% CI 27 to 53) and this was also maintained beyond the intervention period (MD 40. m, 95% CI 24 to 55). There was no immediate, statistically significant difference between treadmill training and overground training in terms of walking speed (MD 0.05. m/s, 95% CI 0.12 to 0.21) or distance (MD -6. m, 95% CI -45 to 33). Conclusion: This systematic review provides evidence that, for people with stroke who can walk, treadmill training without body weight support results in faster walking speed and greater distance than no intervention/ non-walking intervention and the benefit is maintained beyond the training period.