Generative AI and Evidentiary Displacement in Second Language Writing: A Critical Integrative Review of Feedback, Composing Processes, and Assessment Validity

Abstract

Generative artificial intelligence (GenAI) has become embedded in second language (L2) writing not only as a source of correction, but also as a participant in planning, drafting, translating, revising, evaluating, and genre modeling. This critical integrative review argues that the central pedagogical problem raised by GenAI is evidentiary rather than merely technological. In much classroom and assessment practice, learner texts continue to function as privileged evidence for engagement, authorship, composing decisions, and writing ability. GenAI unsettles this relation because it can produce ideas, structure, language, feedback, and evaluation from minimal input while leaving no reliable trace in the submitted text. This review conceptualizes this problem as evidentiary displacement: evidence of learning is not destroyed, but redistributed across less visible processes of prompting, selecting, evaluating, revising, rejecting, and integrating machine output. Drawing on post-2022 work on GenAI-mediated L2 writing and longer-standing scholarship on feedback engagement, composing processes, and assessment validity, the review examines displacement across three sites: feedback, composing, and assessment. It argues for selective process evidence—revision rationales, decision-point annotations, feedback engagement records, and learners’ accounts of their choices—while also cautioning that documentation can increase workload, reproduce inequities, invite surveillance, and become strategically performed rather than pedagogically meaningful.