While numerous studies of robot-assisted language learning (RALL) for English-as-a-foreign-language (EFL) learners' language skill development have been done, a comprehensive and theoretically-driven meta-analysis on its effects is still in paucity. To fill the gap, drawing on Activity Theory (AT), this study reported a meta-analysis from 47 independent studies out of 29 literature samples involving 1791 EFL learners on RALL for language skill development published during 2004-2023. The results indicated that the overall effect size was g = 0.69, 95% CI [0.49, 0.90], suggesting that RALL outperforms non-RALL conditions. In addition, educational levels and intervention durations were found to be significant moderators. Based on the results, implications for practice were discussed.