One way to write an AI policy is not to begin with detection, prohibition, or tool lists. Instead, begin with a central expectation: students must be able to differentiate themselves from AI.
That means their submitted work should show signs of human judgment that are specific to the course: close engagement with assigned materials, situated interpretation, accountable evidence, personal revision choices, and claims that do more than reproduce the smooth middle of what a chatbot might say.
This policy allows some AI use, but makes students responsible for ensuring that the final submission demonstrates their own learning. If a submitted assignment is so generic, detached from course materials, or machine-like that the instructor cannot identify the student’s contribution, the student may be asked to revise or redo the work.
It works best when paired with assignment design that tells students what “differentiation” looks like in practice. In a humanities course, that might mean:
The advantage of this framing is that it keeps the focus on learning rather than surveillance. The risk is that it gives the instructor interpretive authority over what counts as sufficiently “student-like,” which can feel subjective unless the policy is made concrete through examples, rubrics, and chances to revise.
Possible syllabus language
You may use AI tools in this course when they help you brainstorm, revise, test ideas, or better understand difficult material, unless an assignment says otherwise. But you are responsible for making sure that your submitted work clearly reflects your own thinking.
Your work should differentiate you from AI. That means it should show your engagement with course materials, your judgment about evidence, your interpretive choices, and your ability to explain why your claims matter. If I cannot tell where your thinking enters the assignment, I may ask you to revise or redo the work.
Using AI is not a substitute for reading, thinking, drafting, or learning. You remain responsible for the accuracy, integrity, and intellectual substance of anything you submit.
For courses where students are likely to use AI heavily, the policy can add a short disclosure requirement. This makes the standard easier to apply because students have to explain the boundary between assistance and substitution.
Optional AI use note
If AI meaningfully shaped your work, include a brief AI use note at the end of the assignment. In 3-5 sentences, explain what tool you used, what you asked it to do, what you accepted or rejected, and what parts of the final submission represent your own thinking.
This sketch can sit alongside several other kinds of AI policy language. Each one teaches a slightly different lesson.
AI as a citable collaborator: Students can use AI, but must describe the interaction. This works well when the goal is process transparency.
Policy move: process transparency
You may use AI tools as part of your working process, but any meaningful use must be acknowledged in a short note. Name the tool, describe how you used it, and explain how you evaluated or changed its output before submitting your work.
AI use depends on the learning goal: AI permissions change by assignment. This works well in courses where some tasks are about skill practice and others are about critique or experimentation.
Policy move: assignment-specific rules
AI use is governed by the purpose of each assignment. When the goal is practice, close reading, memory, or original analysis, AI may be restricted. When the goal is comparison, critique, revision, or workflow experimentation, AI may be permitted or required. Each assignment will specify what forms of AI use are appropriate.
No undisclosed substitution of labor: The policy focuses on whether AI has replaced the work the assignment exists to develop. This works well when instructors want a clear ethical line without pretending they can police every tool.
Policy move: no substitution
You may not submit AI-generated work as a substitute for your own reading, thinking, analysis, or writing. Using AI to avoid the central intellectual task of an assignment is not acceptable, even if the final text has been edited. When in doubt, ask whether the tool is helping you do the work or doing the work in your place.
This policy gives students a practical standard: their work has to show where they enter the conversation. That can be more useful than a binary permission statement because it teaches students that AI use is not just a compliance issue. It is an authorship issue, an evidence issue, and a learning issue.
It also gives instructors a way to respond to suspicious or generic work without turning every case into a forensic investigation. “Redo this because I cannot assess your thinking” is often a better pedagogical response than “prove you did not use AI.” It preserves the instructor’s responsibility to assess learning while giving the student a route back into the assignment.
Subjectivity is the hard part
A policy built around differentiating student thinking from AI output can be powerful, but it needs examples. Without examples, students may experience the standard as vague or arbitrary, especially if they are still learning what strong disciplinary thinking looks like.
The policy works best when students see concrete contrasts: a generic AI-shaped paragraph beside a paragraph that makes a risky, specific, evidence-based move. Otherwise, “differentiate yourself” can sound like a vibe rather than an assessable expectation.
It is also worth naming the fairness problem. Some students write in polished, generic academic prose because they have been trained to do exactly that. Others may sound unlike themselves because they are multilingual writers, anxious writers, or writers still trying on disciplinary voice. A redo policy should be framed as an opportunity to produce assessable evidence of learning, not as an automatic accusation.
I would not rely on the syllabus statement alone. The policy needs to appear again inside assignments, especially in the first few weeks, with language specific to that task:
I would also build in a low-stakes practice version before using the policy on a major assignment. Students could compare an AI-generated response with a stronger student response and identify where the human thinking becomes visible.
These are the questions that would make the policy stronger before publishing it as a polished sketch: