District policies take a long time to write. Department policies take almost as long. Meanwhile, students walk into your classroom on the first day of the term and ask, reasonably, whether they can use ChatGPT to brainstorm.
Quick Answer
A working classroom AI policy needs four sections: permitted uses by assignment type, disclosure requirement, disputes and evidence, and a review date. Distribute on day one, get student acknowledgment, and revisit at midterm. Escalate to department policy when consistency across sections becomes a problem.
The full guide walks through why the individual layer matters, what to include, sample language for each section, and how to roll it out without turning day one into a lecture.
Why Individual Classroom Policies Matter
An individual classroom policy is not a substitute for institutional guidance. It is the layer of the policy stack you actually control. It also happens to be the layer students read, because it is attached to the syllabus they signed up for.
We have watched teachers spend a semester waiting for their department to align. The students did not wait. They made assumptions. A one-page classroom policy prevents most of the ambiguity that becomes an integrity meeting three months later.
A one-page classroom policy prevents most of the ambiguity that becomes an integrity meeting three months later.
The Working Educators editors
The Four Sections Every Policy Needs
Section one: permitted uses by assignment type. Not one blanket rule for the whole term. Specify, per assignment category, whether AI is permitted, restricted (only for brainstorming or grammar), or prohibited. Vague policies collapse under real cases.
Section two: disclosure requirement. If a student uses a permitted AI tool, they must disclose it. Specify the format: a footnote, a header line, an appendix. Make it easy to comply with. Our guide to citing ChatGPT in academic work lays out language you can adapt.
Section three: disputes and evidence. State plainly that a detector flag is not evidence sufficient for a sanction on its own. Describe what you will do if you suspect an integrity violation: request drafts, ask for a brief conversation about the writing process, and document your reasoning.
Section four: review date. Put a specific date on the policy, no later than one term out. AI capability moves fast enough that any policy written today will need revision within a year.
Sample Language You Can Adapt
Permitted uses: “In this course, generative AI tools such as ChatGPT and Claude may be used for brainstorming and outlining on major essays, and for grammar review on any assignment. AI-generated text pasted into a submission is not permitted on any graded work in this course.”
Disclosure: “Any use of a generative AI tool on a graded assignment must be disclosed in a footnote at the end of the submission, naming the tool and the purpose. Failure to disclose permitted use is treated as a policy violation, separate from any question of AI-generated text.”
Disputes: “AI detection reports are informational only. Any concern about AI-generated text will be resolved by a conversation between the instructor and the student, and where necessary by review of drafts and process artifacts. No sanction will be issued on the basis of a detector percentage alone.”
Review date: “This policy is effective for the current term and will be reviewed and republished no later than [date one term forward].”
How to Roll It Out
Distribute the policy on day one. Read it aloud in the first class session. Post it inside your LMS course home page. Attach it to your syllabus and reference it in the first assignment prompt.
Ask students to acknowledge it, in writing or via an LMS quiz. This creates a paper trail if a dispute later requires you to show the student was on notice. It is not paranoid. It is procedural.
Bring the policy up again at midterm. AI tools that did not exist in September may be dominant by November. If a new tool changes the picture, say so publicly, update the policy, and note the revision date.
When to Escalate to Department Policy
An individual classroom policy is a stopgap. If three teachers in the same department are running three different policies, students get confused, integrity meetings get inconsistent, and the department will eventually be forced to align.
Escalate when at least two of the following are true. First, you are grading writing assignments that also appear in colleagues’ courses. Second, students are asking why the rules differ across sections. Third, an integrity meeting has surfaced ambiguity that a shared policy would have prevented.
Our guide to writing an AI policy faculty will actually follow is written for the department level. The core lesson is that departmental policies survive longer when they are short, specific to real assignment types, and built by the faculty who will enforce them rather than handed down from above.
Frequently Asked Questions
Can I just adopt my department’s policy?▼
If it exists and is specific, yes. Most department-level policies are still generic, which is why a classroom-specific supplement helps.
Should I ban AI outright?▼
That is a values call. If you ban it, be specific about what “AI” means in your policy and prepared to enforce the ban across many tools students already have.
Do I need student signatures?▼
Some form of acknowledgment helps in a dispute. An LMS quiz item is usually enough.
How often should I revise?▼
At least once per academic year. In practice, we revise ours at midterm when the tool landscape shifts.
What if my institution provides no guidance?▼
Then your policy is the operative document for your students. Keep it short, keep it defensible, and put a review date on it.
The Bottom Line
The most durable AI policy in a school is the one the teacher wrote and can enforce on Monday. Do not wait for a memo from the district office.
Write one page, cover the four sections, and revisit at midterm. When your section-level policy conflicts with a colleague’s often enough to confuse students, that is the signal to help the department align around a shared version.
