Abstract

Reading while listening has been shown to facilitate comprehension and vocabulary development in second language (L2) contexts, yet little is known about how learners’ attitudes toward Artificial Intelligence (AI) versus human voices may influence incidental vocabulary learning and online reading behaviour during reading while listening. Addressing this gap, the present study examined whether L2 English learners’ perceived narrator type (AI‑perceived vs. human‑perceived narration) affected (a) incidental acquisition of target vocabulary items and (b) eye‑movement patterns during reading while listening. Sixty university learners completed a vocabulary pre- and post-test measuring form-recognition and meaning-recall, two reading while listening activities with eye tracking, comprehension checks, and an AI perception questionnaire. Multilevel logistic regression analyses revealed robust vocabulary gains from pre-test to post-test; however, perceived narrator type did not significantly predict form-recognition or meaning-recall outcomes. Learning-focused models showed that dwell time was the strongest predictor of form-recognition learning, whereas meaning-recall was more strongly associated with frequency of exposure and comprehension. Eye‑tracking analyses further showed that perceived narrator type did not modulate word fixation probability, gaze synchronisation with narration, or temporal alignment of first fixations. Readers typically progressed ahead of the audio stream, confirming a reading‑ahead profile characteristic of reading while listening. Overall, the findings indicate that incidental vocabulary learning during reading while listening is shaped by attentional allocation and frequency of exposure rather than by narrator framing. These results suggest that AI‑generated voices can be used in reading while listening without compromising lexical learning, with implications for pedagogy, multimedia material design, and future research employing AI-generated stimuli.