Testing Krashen’s input hypothesis with AI: a mixed-methods study on reading input and oral proficiency in EFL

This study investigated the impact of AI-generated graded reading materials on the oral proficiency of adult EFL learners in a six-month intervention. Ninety participants generated weekly texts using proficiency-aligned prompts and were assessed through pre- and post-intervention ACTFL Oral Proficiency Interviews, complemented by learner reflective journals. Quantitative results suggested significant proficiency gains across all initial levels, while thematic analysis of journals highlighted perceived benefits in vocabulary development, autonomy, and fluency. Together, these findings provide preliminary evidence consistent with Krashen’s Input Hypothesis, while also linking AI-mediated reading to broader frameworks of scaffolding, vocabulary acquisition, and cognitive load management. At the same time, important limitations must be noted. The study relied on a single non-certified rater, lacked a control group, and did not systematically monitor the linguistic properties of AI-generated texts. Attrition was concentrated among Novice High learners, raising concerns about bias in proficiency outcomes. These constraints require cautious interpretation, and the results should be viewed as suggestive rather than definitive. Despite these limitations, the study contributes to current discussions on AI in language education by illustrating how generative tools can provide scalable, proficiency-aligned input. It offers preliminary insights into the potential of AI-mediated reading to support oral proficiency development, while underscoring the need for more rigorous designs in future research.