Co-Constructing AI Boundaries: Agency, Judgment, and Ethical Literacy in AI-Mediated Meaning-Making

As generative artificial intelligence (AI) becomes embedded in everyday literacy practices, pre-service teachers (PSTs) must decide what work remains distinctly human in AI-mediated meaning-making. This qualitative case study investigates how PSTs enact boundary work with generative AI during a semester-long literacy ethnography. Drawing on Nexus Analysis and digital trace data from Google's NotebookLM, I analyze moments of agency, refusal, and ethical judgment within the human–AI evaluative loop. Findings identify two contrasting interaction profiles, the Orchestrator and the Outsourcer. Orchestrators retained epistemic authority by repeatedly constraining, correcting, and re-authoring the AI system's output, while Outsourcers delegated organizational and interpretive work to the system early and accepted its framing with little revision. These differences in how PSTs worked with the AI system were visible only through process-level analysis of interaction traces, as the final artifacts appeared deceptively similar. The study reconceptualizes human-in-the-loop as an interactional literacy practice rather than a technical safeguard. Implications highlight the need for teacher education to design “loops worth living in” that normalize productive friction and professional refusal, ensuring AI amplifies rather than replaces human judgment.