Reliability of AI in the domain of literary literacy and literary text interpretation: an empirical study

Abstract The article presents the results of an experimental study in the field of L1 education that used literary literacy and the interpretation of literary texts as examples to examine the reliability of generative AI in helping learners in literature classes to overcome complex challenges. Due to the ambiguity of literary texts, the reliability of AI faces particular challenges here. Based on an empirically verified model of literary literacy, two chatbots considered particularly powerful—ChatGPT-5 from Open AI and Claude Sonnet-4.5 from Anthropic—were systematically tested. After explaining the study design, the initial results of the experimental study are presented and discussed.