Evaluating AI feedback: teacher expertise and hidden labour in writing instruction

Generative AI (GenAI) is increasingly used to comment on student writing in both L1 and additional-language classrooms. Although often presented as a way to reduce feedback workload, it also creates a new professional task: teachers must judge whether AI-generated feedback is accurate, developmentally appropriate, and suitable for the task, genre, and institutional context. This conceptual article names that work ecological evaluative labour (EEL). Drawing primarily on L2 writing research and broader scholarship on writing pedagogy, it develops an argument relevant across L1 and additional-language instruction. The framework distinguishes three interrelated domains of teacher judgement: accuracy, developmental appropriateness, and normative-contextual fit. It also proposes an expertise asymmetry: experienced writing teachers may bear greater evaluative burdens because they notice mismatches that less experienced teachers may overlook. Because GenAI outputs vary across prompts, contexts, and model configurations, this work may become more efficient but cannot be fully routinised. An illustrative writing case and teacher comments clarify, rather than empirically test, the framework. The article concludes that responsible GenAI use should be recognised as a form of professional judgement and workload, with implications for novice teachers, experienced teachers, and institutional leaders.