AI as a Learning Partner: Error Analysis, Text Revision, and Student Perceptions in German Writing

Abstract. This study investigates the role of artificial intelligence (AI), specifically Deepseek, as a learning partner in A2-level German writing (Schreiben). The research focuses on error analysis, revision processes, and students’ perceptions of using Deepseek in writing tasks. Participants were second-semester students of the German Language Education Program at State University of Jakarta. A descriptive qualitative method was employed, drawing on student texts before and after correction, Deepseek’s feedback, and responses to an open-ended questionnaire on students’ experiences. The procedure includes six steps: (1) student produced a text on a predetermined topic, (2) the text was corrected by Deepseek using structured prompt, (3) student revised their texts based on AI feedback, (4) Students were devided into groups and discuss the result of DeepSeek correction, (5) each group presented the result of their discussion, and (6) student perceptions were gathered through questionaries. Findings indicate that the most frequent errors concerned grammar, particularly case usage (Akkusativ, Dativ), prepositions, verb conjugation, articles, and word choice. Students expressed predominantly positive perceptions, noting that AI fosters writing autonomy, error awareness, and provides immediate feedback. The study concludes that AI can serve as an effective learning partner in foreign language writing, particularly German, by improving the quality of student texts and fostering learner autonomy. Keywords: AI Feedback, Deepseek, Schreiben, Error Analysis