Background: Generative artificial intelligence (GenAI) chatbots have shown great application potential and research value in the field of language learning, especially in second language acquisition (SLA). Objectives: This study aims to validate the value of GenAI chatbots for SLA and elucidate which moderating variables affect learning outcomes. Methods: The method of meta-analysis was adopted to conduct this study. This study synthesised 41 experimental or quasi-experimental studies on the application of GenAI chatbots in SLA from 2023 to date and extracted 48 independent effect sizes from them, with a total of 3515 participants. Results and Conclusions: The results showed that GenAI chatbots had a moderate to large positive impact on SLA (ES = 0.576, 95% CI = [0.385-0.768], p < 0.001). In addition, based on the theoretical framework, this study further explored the moderating effect of GenAI chatbots on SLA from the dimensions of adopters of innovation, innovation, time process, diffusion channels and social system. To be specific, there were no inter-group differences in educational stage and learning environment, and there were intra-group differences in other moderating variables. Among them, the first language (L1) represented by Indonesian, the second language (L2) other than English, the GenAI chatbot represented by ChatGPT, vocabulary as the learning objective of language, and the intervention time of 1-7 days had a large moderating effect. Additional implications for research and practice are discussed.