Key points are not available for this paper at this time.
Since its emergence, generative AI has significantly impacted various fields, including English language education. For example, numerous academic studies have investigated the accuracy of grammar correction, the evaluation of writing, and the dynamics of user interaction. However, there has been insufficient investigation into whether texts generated by such AI appropriately align with CEFR levels. This study addresses this gap by exploring the applicability of generative AI to CEFR levels. Multiple texts were generated using ChatGPT-4 with specified CEFR levels and analyzed using a vocabulary level analyzer (CVLA). The findings revealed discrepancies between AI-generated texts and textbook standards, significant divergences between levels below B1 and above B2, and a noticeable topic bias. Although the texts generated by the AI appear to differ by level on the surface, caution is warranted when applying them to CEFR levels.
Satoru Uchida (2025) studied this question.