Nan Bi

The relevance of the study is driven by the need to revise the teacher's methodological toolkit in the context of the rapid integration of generative AI models into the educational process, particularly into teaching Russian scientific written speech to foreign students. The need to overcome the key contradiction between the ability of generative AI to instantly produce ready-made text and the need for step-by-step development of the skill of independently formulating scientific thought in non-native speakers is of particular importance. The aim of the study is to identify the areas of AI application in teaching Russian scientific writing to non-native speakers and to determine the conditions under which AI becomes a means of developing skills rather than replacing them. A comprehensive approach is aimed at identifying the practices, difficulties and deficits of non-native speakers with B1 level of Russian as a foreign language when using AI in working with scientific texts. The analysis focused on four aspects: popular scenarios of AI use, vague perceptions of academic ethics, the gap between students' self-assessment of their skills and their actual behaviour, and the contradiction between awareness of risks and the lack of verification strategies. The methodological framework includes a theoretical analysis of research on the use of AI in teaching scientific writing, a questionnaire survey of non-native speakers, a qualitative analysis of their detailed responses, and pedagogical observation of the completion of tasks on writing secondary scientific texts. The results of the study demonstrate that non-native speakers widely use AI for translating and analysing scientific texts, do not consider this a violation of ethics, do not cross-check the responses and encounter hallucinations. While being aware of the risk of degradation of their own skills, they have no protection strategies. The paradox is that this awareness does not translate into verification practice – students know about the danger but do not check the information and do not change their behaviour. The novelty lies in redefining the role of AI from a generator of ready-made text to an editing tool, in identifying two priority skills, namely competent formulation of queries to AI and critical verification of the information obtained, as well as in identifying a methodological risk consisting in students' use of generated false sources without verification. The theoretical significance lies in the development of scientific understanding of the boundaries of AI applicability in teaching written scientific discourse and in clarifying the concept of "methodological expediency" in relation to the automated formation of scientific discourse literacy. The practical significance lies in the fact that data on students' deficits when working with AI allow the teacher to adjust instruction and substantiate the need to develop students' skills in prompting, verification and observance of academic ethics.