These Are Not My Words: Writer‐AI Alignment, Template Attraction, and Verification in AI‐Mediated EFL Writing at a Jordanian Public University
TL;DR
The study shows that critical AI literacy in L2 writing is enacted through revision, verification, and ownership decisions, not through attitudes alone.
Abstract
Generative artificial intelligence (GAI) is now widely used in L2 writing, but much existing research still relies on learner attitudes rather than traceable evidence of writing behaviour. This study examines how Jordanian EFL writers judged, used, revised, verified, and sometimes rejected AI‐generated language during actual composing. It asks how students evaluated the alignment between AI wording and their intended meaning, voice, genre purpose, and evidential responsibility; what uptake pathways they followed when working with AI output; and how verification shaped revision depth and authorial ownership. The study draws on a qualitative‐led multiple‐case design with embedded descriptive episode analysis. Thirty second‐year English majors at a Jordanian public university completed three course‐based assignments: a personal narrative, a source‐based argument essay, and a public‐facing opinion piece. The corpus included AI chat logs, draft pairs, writer memos, teacher feedback, verification logs, screen‐session logs, stimulated recalls, diagnostic samples, and descriptive score records. Analysis centred on 196 traceable revision episodes linking prompts, AI responses, writer decisions, and textual changes. Four patterns were identified: fluent but wrong, the comfort of ready‐made English, voice under pressure, and trust must be earned. Students used AI productively for structure, option generation, and sentence reshaping, but resisted wording that generalized local experience, imposed a ready‐made stance, or exceeded available evidence. The study shows that critical AI literacy in L2 writing is enacted through revision, verification, and ownership decisions, not through attitudes alone.