OBJECTIVES: To accumulate evidence that obstacle avoidance training alone is effective in improving the locomotor ability of individuals with stroke. DESIGN: Systematic review and meta-analysis. SETTING: MEDLINE, EMBASE, CENTRAL, ICTRP and PEDro were searched for related information until December 2018. Two independent reviewers extracted data. Outcome measurement data were subjected to meta-analyses using random-effects models. Data syntheses were conducted using RevMan V.5.3, and the certainty of evidence was determined using the Grading of Recommendations Assessment, Development, and Evaluation approach. PARTICIPANTS: Participants with various types and phases of stroke were included. INTERVENTION: The usual gait training including obstacle avoidance training (interventions of any type, intensity, duration and frequency). PRIMARY AND SECONDARY OUTCOME MEASURES: Primary outcomes were gait speed, composite gait ability and objective balance ability. Secondary outcomes were subjective balance ability, gait endurance and fall incidence. RESULTS: Two randomised controlled trials with a total of 49 participants were used as data sources for this study. The obstacle avoidance training (training) group had lower gait speed than the control group (mean difference (MD) 0.03, 95% CI -0.11 to 0.16, p=0.51). Further, the certainty of evidence was very low. The subjective balance ability (Activities-specific Balance Confidence scale) was not significantly different between the training and control groups (MD 6.65, 95% CI -7.59 to 20.89, p=0.36), and it showed very low certainty of evidence. CONCLUSIONS: Obstacle avoidance training may have little or no effect on individuals with stroke. The failure to find the effectiveness of obstacle avoidance training alone is possibly attributable to the insufficient amount of training in the intervention and the lack of well-designed studies that measured relevant outcomes. PROSPERO REGISTRATION NUMBER: CRD42017060691.