Exploring ChatGPT's Role in Developing Productive Language Skills: A Systematic Review

Abstract

Although the interactive features of ChatGPT offer substantial benefits for language education, its integration presents complex pedagogical and ethical challenges. This systematic review critically evaluates the capabilities and limitations of ChatGPT in enhancing expressive language capabilities, specifically speaking and writing. Adhering to PRISMA guidelines, a rigorous search of the ERIC database and a forward snowballing technique via Google Scholar yielded an initial pool of 1,124 records, culminating in the final selection of 22 empirical studies for in-depth analysis. Grounded in Vygotsky's Sociocultural Theory and the SAMR integration model, the findings conceptualize ChatGPT as a virtual ``More Knowledgeable Other'' that provides dynamic cognitive scaffolding. The synthesis indicates that ChatGPT substantially improves writing proficiency by delivering immediate, personalized feedback, assisting with structural organization, and alleviating academic stress. In conversational contexts, the tool provides a safe, non-judgmental platform that lowers the affective filter and encourages autonomous speaking practice. However, critical engagement with the literature reveals profound limitations. While the AI excels at the structural mechanics of language production, it currently falters at the socio-pragmatic realities of human connection, lacking emotional depth in conversational dialogues. Furthermore, the review identifies critical risks, including AI hallucinations, the degradation of authentic academic voice, and a paradox wherein increased learner autonomy inadvertently leads to AI dependency and ``cognitive outsourcing.'' Ultimately, this paper argues for a paradigm shift in language pedagogy, emphasizing the urgent need to integrate critical AI literacy into curricula so that learners utilize generative tools as temporary scaffolds rather than permanent cognitive crutches.