ChatGPT in Undergraduate English Language Majors: Benefits and Challenges for Writing and Speaking Proficiency

Abstract

Large language models such as ChatGPT are reshaping undergraduate English language education. This systematic review synthesizes benefits and challenges for developing writing and speaking proficiency among English language majors. Following PRISMA-2020/PRISMA-S and SPAR-4-SLR, we searched Scopus, Web of Science, ERIC, EBSCO and SSRN, and analyzed 31 of the 708 empirical and review papers published. Findings converge on writing gains when ChatGPT is embedded in scaffolded, human-in-the-loop workflows that emphasize pre-writing, drafting, and revision. Effects are strongest for accuracy, coherence, and argument quality, and when teacher or peer moderation, prompt scaffolds, and transparent rubrics are present. For speaking, learners benefit through low-stakes practice, rehearsal and anxiety reduction; however, robust measurement lags behind, with scarce CEFR-aligned rubrics, limited voice-mode instrumentation, and few validated acoustic indicators. Integrity and equity remain central. Text-only AI detectors are brittle and sometimes unfair to non-native writers. Institutions should pivot to process-anchored assessment - combining prompt logs, version histories, and brief viva voce - to evidence authorship while preserving learning value. A forward research agenda is proposed, including a minimum reporting toolkit for speaking measurement (CEFR, ASR features, and ICC/κ) and multi-site randomized trials.