Purpose This systematic review and meta-analysis aimed to assess the efficacy of non-invasive interventions for reducing phantom limb pain (PLP), evaluate the certainty of evidence using GRADE, and explore potential moderators through meta-regression.Materials and methods Searches were conducted in PubMed, Scopus, Embase, Web of Science, and Cochrane to identify randomized controlled trials involving adults with PLP. Random-effects models were used for the meta-analysis. Certainty of evidence was assessed using GRADE, and meta-regression analyses were performed to explore potential moderators.Results Twenty-six studies met the inclusion criteria. Mirror therapy (MT) produced small but significant reductions in pain compared with sensorimotor therapy, although no significant effects were observed versus sham interventions. Graded motor imagery (GMI), virtual and augmented reality (VR/AR), and repetitive transcranial magnetic stimulation (rTMS) demonstrated consistent improvements in PLP, whereas electromagnetic liners showed no benefit. Certainty of evidence ranged from low to moderate. Meta-regression did not identify significant moderators (p > 0.22), suggesting relatively stable treatment effects across demographic and methodological variables.Conclusions Interventions targeting central sensorimotor mechanisms, particularly MT, GMI, VR/AR, and rTMS, appear to be effective in reducing PLP; however, the overall quality of evidence remains limited.