<p>As generative Artificial Intelligence (GenAI) tools become increasingly embedded in academic writing practices, questions of authorship, integrity, and accountability require new assessment approaches. This study examines teacher-led interviews as a mediational assessment practice, informed by an integrative cultural–historical activity theory (iCHAT) perspective that emphasizes how reasoning, responsibility, and authorship develop through guided social interaction. Teacher-led interviews were designed to prompt students to explain and justify their research writing decisions, including how they engaged with GenAI support. Reflection data were collected from 24 Master’s-level English as a Foreign Language (EFL) students, who were asked to participate in short teacher-led interviews about their research proposal submissions as part of the unit assessment. Thematic analysis indicates that the interviews supported the emergence of an ethical orientation toward academic work, expressed through cognitive, emotional, and agentive developments. Students engaged more critically with ideas and AI-generated suggestions (cognitive), negotiated feelings of anxiety, fairness, and recognition (emotional), and adopted more proactive and self-regulated preparation strategies (agentive). These outcomes appeared to be shaped by how students experienced the interview (perezhivanie), suggesting that the interview reoriented the object and motive of the writing task from producing text to being able to account for one’s reasoning and decisions. The study proposes a cultural-historical framework conceptualizing teacher-led interviews as a mediational space where dialogic accountability supports reflective, ethical, and principled engagement with GenAI. Implications highlight the need for assessment practices that cultivate development through guided explanation rather than rely solely on detection or surveillance.</p>