TL;DR
The results indicate that ChatGPT significantly improves language quality in six of the seven assessed indicators, with these improvements occurring consistently across all proficiency levels.
This study explores how multilingual 9th-grade students use generative AI as a writing assistant in German (L1/L2) and English (FL) writing contexts by focusing on strategy use, cross-linguistic transfer, and proficiency-related variation. Drawing on a mixed-methods design that combines qualitative and quantitative data, the analysis suggests that generative AI may introduce a new mediating layer in the writing process, enabling learners to externalize cognitive subtasks such as planning, drafting, and revising. Five main AI writing strategies were identified, ranging from planning and linguistic optimization to metacognitive reflection. Overall, findings indicate patterns consistent with cross-linguistic transfer of AI writing strategies in this exploratory sample, highlighting AI’s potential to foster metacognitive awareness and cross-linguistic strategy development in multilingual writing. However, proficiency appears to moderate these patterns: More proficient (Level B2) learners of English as a foreign language show greater strategic stability in their AI use, whereas lower-proficiency learners engage with AI in a more situational and cautious manner, a pattern that is also reflected in two illustrative learner cases.
This empirical corpus-based TESOL study examines mode and proficiency effects in EFL learner production through a planned comparison of written essays and spoken monologues in the International Corpus Network of Asian Learners of English (ICNALE). The study focuses on how lexical, grammatical, and selected discourse features vary across written and spoken learner output, and how these patterns interact with proficiency level. Using a quantitative comparative corpus design, the study will analyze measures such as word count, lexical density, lexical diversity, academic vocabulary use, syntactic complexity, discourse markers, stance markers, and selected fluency-related transcript features where available. At this draft stage, the final ICNALE sample, feature counts, statistical outputs, and figures have not yet been supplied; therefore, all numerical findings are marked for author completion and verification. The manuscript is structured to allow the author to insert verified values after registering for ICNALE access, selecting comparable prompts, cleaning the dataset, and completing the analysis. The expected contribution is a mode-sensitive account of learner production that may inform integrated TESOL writing and speaking instruction, proficiency-sensitive assessment, and corpus-informed materials development. The paper differs from corpus-literacy or teacher-education review work by focusing on empirical learner-language evidence rather than teachers’ corpus-use knowledge.
It is suggested that AI integration alone does not guarantee improved writing outcomes without corresponding foundational skill development, as students lack the foundational knowledge necessary to utilize AI tools effectively.
It is argued that the thoughtful integration of AI feedback with teacher feedback, grounded in pedagogical principles and human judgment, can significantly enhance L2 writing instruction.
The analysis reveals that reflective essays written by students highly dependent on artificial intelligence display a distinct “conceptual emptiness” in which surface-level fluency coexists with inner superficiality, and concrete linguistic criteria enabling educators to perceive patterns of AI intervention in student assignments is offered.
We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.
