The new era of generative artificial intelligence has sparked the blossoming academic fireworks in the realm of education and information technologies. Driven by natural language processing (NLP), automated writing evaluation (AWE) tools become a ubiquitous practice in intelligent computer-assisted language learning (CALL) environments. Based on the self-set corpus of the plain text file encompassing 1524 documents from the Web of Science core collection, the current study adopts quantitative and qualitative methods and integrates bibliometric, scientometric, and meta-analytic approaches aiming to comprehensively review automated writing evaluation (AWE) over fifteen years from 2008 to 2023. Feedback literacy is the theoretical framework of automated written corrective feedback (AWCF). Through VOSviewer, this study bibliographically visualized AWE-relevant keywords, documents, authors, organizations, and regions at a macro level. Science mapping analysis (SMA), mapping knowledge domain (MKD), and author co-citation analysis (ACA) are the theoretical foundations of visualization on VOSviewer. Through Stata/SE 16 and SPSS 29, this study meta-analytically investigated moderator effects of various AWE tools, feedback types, intervention duration, target language learners, educational levels, genres of writing, regions, document types, and publication year at a micro level. It is concluded that AWE tools could facilitate writing proficiency at a statistical significance level (SMD = 0.422, p < 0.001) based on 29 experimental studies. The findings illuminate future research directions and provide heuristic implications for practitioners, researchers, and AWE technology developers.