Background BCI-based lower limb robots (BCI-LLR) represent a novel technology used in neurological rehabilitation for stroke patients. However, the effectiveness of BCI-LLR compared to traditional rehabilitation in improving lower extremity function and cognition remains a topic of debate. This study aimed to determine whether BCI-based lower limb robots are more effective than traditional rehabilitation for lower limb dysfunction after stroke. Methods A comprehensive search was conducted across multiple databases, including PubMed, Embase, Web of Science, and Cochrane Library. The database retrieval was performed up until November 21, 2025. A systematic review of findings was conducted due to the limited number and the heterogeneity of studies. Results A total of four studies (three RCTs and one single-arm pre-post study) with 117 stroke patients were included. BCI-LLR training significantly improved lower limb function, with one RCT showing significant between-group differences in FMA-L (+4.5 vs. +2.1, p = 0.022), and another RCT demonstrating significant advantages in TUG (−12.03 vs. −4.65 s, p = 0.038) and BBS (+5.50 vs. +3.38, p = 0.042). The single-arm study also reported significant improvements in all lower limb outcomes ( p < 0.01). For cognition, SDMT improvements significantly favored BCI-LLR in two RCTs ( p = 0.036 and p = 0.013) and in the single-arm study ( p = 0.047). MoCA showed improvement only in the single-arm study ( p = 0.044). The attention index increased significantly in three studies ( p < 0.01). Neurophysiological benefits were reported in one RCT ( p < 0.05). Conclusion Preliminary evidence suggests BCI-LLR may improve lower limb function (FMA, BBS, TUG) and information processing speed (SDMT). However, effects on global cognition (MoCA) remain unclear, and sample sizes are critically small. Systematic review registration This study was registered on the international system evaluation registration platform PROSPERO (CRD420251232191).