The relevance of the study stems from the growing need to introduce artificial intelligence into the educational process, particularly in teaching non-Latin languages, such as Japanese. The combination of optical character recognition and AI-based analysis opens up new opportunities to improve the efficiency of writing assessment, the quality of feedback, and the individualisation of learning. This study aimed to investigate the use of ChatGPT to improve the effectiveness of Japanese language teaching through the automated checking of student essays. Methodologically, educators used the OCR function of ChatGPT to digitise handwritten student essays, enabling analysis of grammar, vocabulary, sentence structure, and kanji. A mixed-methods approach was adopted, combining quantitative error analysis with qualitative insights from instructors. The study revealed that ChatGPT effectively detected eight major error categories, with the most common being insufficient kanji usage, overuse of basic sentence structures, and a limited vocabulary range. The tool’s ability to generate instant error summaries and targeted feedback significantly reduced the time educators spent on marking, ensuring consistent and individualised support for learners. However, the study also identified limitations, including occasional inaccuracies in OCR processing, especially for stylised or unclear handwriting. Despite these challenges, the findings suggested that ChatGPT can serve as a valuable complementary tool to traditional teaching methods, enhancing feedback quality, supporting differentiated instruction, and promoting learner autonomy. The findings offer practical guidance for language educators, curriculum designers, and edtech developers seeking to integrate AI into the teaching and assessment of less commonly taught languages, such as Japanese

generative AI; language proficiency evaluation; AI in education; automated feedback; Japanese language pedagogy