One 45-Minute Session, 26 AI-Assisted Drafts

I would reframe the session around one high-leverage skill tied to the actual assignment: evaluating and documenting sources, not citation formatting in isolation. Before the visit, I would ask the teacher for the assignment sheet and three sample student topics, then build a one-page source-evaluation checklist using the students' real bibliography entries. In the session, I would model lateral reading with two sources-a peer-reviewed article and a think-tank blog-and have students evaluate their own sources and write a short source note. MLA format becomes a ten-minute mini-lesson plus a reference guide, because formatting is a lookup skill and does not need the whole period. I would negotiate one follow-up touchpoint, even fifteen minutes in the classroom or an asynchronous video, because a single session cannot change 26 students' research process. The reasoning is transfer: students need practice with authentic sources, not a lecture about punctuation.

I would explain that AI detectors are not reliable evidence: they produce false positives, especially for neurodivergent students and multilingual writers, and a positive result cannot support discipline. I would redirect the teacher toward process-based evidence-annotated bibliographies, drafts, reflection memos, in-class writing, and brief oral defenses-because those show whether students can explain and revise their work. For the fabricated citations, I would treat that as an information-literacy problem to teach, not only a cheating problem: students need to verify that a source exists, locate it, and read enough to cite it accurately. I would recommend redesigning the assignment to reduce temptation: require primary sources, a local angle, or a multimodal component. If the district has an AI policy, follow it; if not, I would propose one with the principal and department chair. My role is to support ethical use and source verification, not to run surveillance.

I would use a short pre/post performance task rather than a satisfaction survey. Students would receive two sources on the same topic-one peer-reviewed article and one advocacy blog-and answer three prompts: Who created this, and what expertise do they have? What evidence is offered, and what is missing? How would you use this source in a paper? I would score with a common rubric (authority, evidence, bias/purpose, contextual fit), and I would run the same task at the start and end of the unit. I would also ask students to audit their own final bibliographies for verifiability. The department chair is skeptical that library instruction affects scores, so I would connect the rubric to the research standards the department already uses and report growth in specific categories, not a single average. Low-stakes, anonymous data can be shared without exposing individual students, and the results give the department a concrete basis for deciding whether to expand collaboration next year.