<p>This study investigates the effects of stimulus modalities (textual, auditory, and audiovisual) and language proficiency (L1 and L2) on bilingual writing production, focusing on writing accuracy and reaction times. Thirty Chinese EFL learners (mean age = 18.13) participated in a within-subjects experiment using a micro-level writing task involving 15 non-cognate animal words. Stimuli were presented via E-Prime 3.0 in randomized modality–language combinations, and writing responses were recorded using a WACOM tablet and electronic pen to capture reaction times. Each participant translated single words from L1 to L2 and vice versa under three stimulus modality conditions. Linear mixed-effects models (LMMs) were used to analyze reaction time and accuracy data, with stimulus modality and language type as fixed effects and participants and items as random effects. The findings reveal significant performance differences across modalities, with textual and audiovisual stimuli yielding higher accuracy and faster reaction times than auditory input. Language proficiency moderated these effects, with L1 outperforming L2 across all conditions. The results underscore the benefits of multimodal input, particularly audiovisual stimuli, for supporting low-proficiency L2 writers. These findings carry practical implications for the design of Computer-Assisted Bilingual Education (CABE) systems, emphasizing the value of adaptive, proficiency-sensitive, and multimodal features in enhancing bilingual writing instruction. Future research could incorporate process-tracing methods such as eye-tracking to complement reaction time measures and more comprehensively capture how different stimulus modalities influence bilingual writing processes.</p>