Every writing instructor I talk to has some version of an AI policy. It might live in the syllabus as a single sentence. It might be a paragraph borrowed from another department. It might be the institution's boilerplate, pasted in without modification.
Most of these policies share two problems. They are either so vague that students cannot tell what's actually permitted, or so absolute that they don't survive contact with how students actually use AI in their daily lives. Neither version serves the instructor, the student, or the course.
This post is about what a good AI policy actually needs to do, what it needs to say, and what it can safely leave out.
What a Policy Is Actually For
Before writing a single word, it helps to be clear about the purpose. An AI policy in a writing course is not primarily an enforcement mechanism. It is a clarity mechanism.
Ambiguity is where academic integrity cases are born. When students don't know what's permitted, some will assume everything is fine. Others will assume everything is prohibited. Most will make their own judgment calls and hope for the best. None of this produces the transparent, documented process that makes AI use educationally meaningful.
A good policy answers three questions before the first assignment is due:
- What AI tools and uses are permitted in this course?
- What must be disclosed, and how?
- What are the consequences for failing to disclose?
That's it. If your policy answers those three questions clearly, it has done its job.
Ambiguity is where academic integrity cases are born. Clarity is where they are prevented.
The Five Components of a Workable Policy
A syllabus-ready AI policy doesn't need to be long. The ARWI AI Transparency Policy template runs to about one page and covers everything an instructor needs. Here's what it includes and why each component matters.
A clear statement of permitted use
This is the most important sentence in the policy. It answers the question students will actually ask: can I use AI or not? The answer should be specific enough to guide behavior, not just general enough to sound official.
Vague: "AI use may be permitted in some contexts at the instructor's discretion."
Clear: "You may use AI tools for brainstorming, research assistance, and feedback on drafts. You may not submit AI-generated text as your own writing without explicit attribution and documentation."
A disclosure requirement with a specific mechanism
Telling students to disclose AI use is not enough if you don't tell them how. The Process Log is the disclosure mechanism in the 4D Model for AI-Resilient Writing™. Students document what tools they used, what prompts they entered, what they kept, and what they independently revised. That documentation lives in a running Google Doc submitted alongside every major draft.
The policy should name the mechanism explicitly so students know what's expected before they start working.
A statement about what counts as your own work
This is where most institutional policies break down. "Your own work" in an AI-present classroom is not the same as "text you typed yourself." A student who used AI to generate an outline, then wrote every sentence independently, produced their own work. A student who pasted AI-generated paragraphs and changed a few words did not.
The policy should define this boundary in plain language, focused on thinking and decision-making rather than just keystrokes.
Consequences for non-disclosure
The policy needs to be honest about what happens when a student uses AI without documenting it. This is not primarily about punishment. It is about making clear that the documentation requirement is not optional. Non-disclosure treated as a minor oversight produces a culture where documentation is treated as a technicality. Non-disclosure treated as an academic integrity violation produces a culture where transparency is taken seriously.
A statement of purpose
The most underused component. A one or two sentence explanation of why the policy exists, framed around learning rather than enforcement, changes how students read everything that follows. Students who understand that the documentation requirement exists to make their thinking visible are more likely to engage authentically than students who see it as a surveillance mechanism.
What to Leave Out
A good policy is as notable for what it doesn't include as for what it does.
Don't list specific tools
Any policy that names specific AI tools will be outdated before the semester ends. ChatGPT, Gemini, Claude, Copilot: the list changes constantly and the underlying question is not which tool a student used but how they used it and whether they documented it. Write the policy around behavior and disclosure, not tool names.
Don't promise detection
Telling students their work will be checked with AI detection tools is a promise you cannot reliably keep. Detection tools produce false positives, flag non-native English writers at disproportionate rates, and are routinely defeated by basic editing. A policy that leans on detection as its enforcement mechanism is built on a foundation that doesn't hold. The documentation requirement is the enforcement mechanism. Let it do that work.
Don't write for the worst-case student
Policies written primarily to catch cheaters tend to read like accusation documents. They create an adversarial tone from the first day of class and signal to honest students that they are not trusted. Write for the majority of students who want to understand the expectations and meet them.
Phrases like "use AI responsibly," "AI must be used ethically," and "unauthorized AI use will result in consequences" sound meaningful but answer none of the three questions a policy needs to answer. Students cannot act on them, and instructors cannot enforce them consistently. Replace every vague phrase with a specific behavior or requirement.
How to Introduce It in Class
The policy is a starting point, not a substitute for a conversation. The most effective AI policies are the ones instructors actually discuss on the first day, explaining not just what the rules are but why they exist and what the course is ultimately trying to develop.
That conversation is the Declare step of the 4D Model for AI-Resilient Writing™. It sets the tone for everything that follows: that this course takes AI seriously, that transparency is expected and supported, and that the goal is not to police tool use but to make student thinking visible.
Students respond differently to a policy introduced as a shared framework than to one handed down as a set of rules. The difference is not just pedagogical. It is practical. Students who understand the purpose of the documentation requirement are more likely to complete it honestly, which produces better evidence of their learning and fewer integrity cases to adjudicate.
A policy introduced as a shared framework produces transparency. A policy handed down as a set of rules produces compliance theater.
The Template
The ARWI AI Transparency Policy template is one of six documents in the free Starter Kit at arwi.ai. It is syllabus-ready, which means it is formatted to drop directly into a course document without reformatting. It covers all five components described above and is designed to be customized: the permitted use statement, the disclosure mechanism, and the consequence language all include bracketed placeholders so you can adjust them to your course context in about ten minutes.
It is also discipline-neutral. The template works in a first-year composition course, a technical writing course, a history seminar, or any other course that assigns writing. The underlying framework is the same regardless of discipline.
Get the AI Transparency Policy Template
Syllabus-ready, customizable, and free. Part of the full ARWI Starter Kit including the Process Log, Reflection Prompts, and Writing Rubric.
Download the Free Starter Kit →