Generative AI and Second Language Vocabulary Processing: A Cognitive Study of Chinese EFL Learners

Abstract

This study investigates the impact of AI-generated texts—specifically narrative, dialogic, and explanatory types—on lexical processing among Chinese English as a Foreign Language (EFL) learners. A mixed-methods approach was employed, combining quantitative measures of lexical decision speed and accuracy, semantic mapping, and retention rates with qualitative insights into strategy use and cognitive experience. Results indicate that dialogic texts significantly enhanced lexical access speed and accuracy, while narrative texts promoted deeper semantic integration and better long-term retention. Explanatory texts, though less effective overall, supported vocabulary learning through structured input, particularly for learners with lower working memory capacity. Regression analyses revealed that both working memory and language proficiency were significant predictors of lexical outcomes, with moderation effects showing that high-working-memory learners benefited more from narrative texts, whereas low-working-memory learners showed greater gains from dialogic texts. These findings highlight the importance of text type selection in vocabulary instruction and underscore the need for adaptive AI-driven systems that tailor content to individual learner profiles. The study contributes to both second language acquisition theory and the pedagogical application of generative artificial intelligence in language learning environments.