Much of the early institutional response to generative AI in writing classrooms centered on detection: catching misuse after a paper was submitted. A meaningful shift is now underway toward a different approach, one that addresses AI use during the drafting process itself through clear, transparent policies about when and how AI tools are acceptable, rather than relying primarily on after-the-fact detection to enforce a blanket prohibition. This shift is showing up in state-level model policies, district guidance, and even student-authored proposals calling for clear tiered frameworks rather than simple bans.

The logic behind this shift is straightforward: students are going to encounter and use AI tools throughout their academic and professional lives regardless of any individual classroom's policy, and teaching them to use these tools transparently and appropriately, disclosing use, understanding when it's appropriate versus when it undermines the purpose of an assignment, builds a more durable and honest skill than a prohibition that mainly teaches students how to avoid getting caught. A number of teacher-developed classroom policies now build this directly into assignment design, requiring students to submit a brief note documenting how and where AI tools were used in their process, alongside the final piece of writing.

Model policies emerging at the state level increasingly reflect this same transparency-first framing: prohibiting AI for fully generating an assignment while explicitly permitting it for brainstorming, editing, or other supporting roles, with disclosure requirements that make a student's process visible rather than trying to catch misuse invisibly after submission.

Building an assignment-level disclosure structure

A workable classroom version of this doesn't require a district-wide policy overhaul to implement individually. A teacher can specify, per assignment, exactly what level of AI assistance is acceptable, none at all for a reflective personal essay, brainstorming and editing support for a research paper, for instance, and require a brief disclosure statement documenting what was used and how. This gives students a clear, low-stakes way to be honest about their process, rather than a system that implicitly rewards concealment.

• Specify AI use expectations clearly per assignment, not as one blanket policy for every writing task
• Require a brief disclosure statement documenting how AI tools, if any, were used in the drafting process
• Distinguish explicitly between AI use for brainstorming or editing versus AI use for generating original content
• Treat disclosure honestly given as a positive, not an admission that triggers automatic suspicion
• Revisit and communicate policy expectations at the start of each major assignment, not just once at the start of the term

What this means for grading

When AI use during drafting is disclosed rather than hidden, grading becomes considerably more straightforward: a teacher can evaluate the final piece of writing on its own merits, with a clear understanding of what kind of assistance shaped the process, rather than trying to retroactively guess at authorship through unreliable detection tools. This approach doesn't eliminate every academic integrity question a classroom faces, but it shifts the default interaction from suspicion to transparency, which tends to produce more honest student behavior over time, not less.

It's worth being clear that this transparency principle applies to teachers using AI-assisted grading tools too. A district or classroom expecting honest disclosure from students about their AI use during drafting has a parallel obligation to be transparent with students and families about AI's role in the grading process, the same human-in-the-loop, teacher-reviews-every-grade principle that responsible AI grading tools are built around.

Where this is headed

The transparency-first approach to student AI use is still relatively new, and best practices are actively evolving as more schools pilot different disclosure structures and compare outcomes. What seems clear already is that a purely detection-based approach, given the well-documented reliability problems with AI writing detectors, was never going to be a durable long-term strategy on its own, and the shift toward teaching honest, disclosed use during the drafting process itself looks like the more sustainable direction for classrooms to build toward.