Building an AI Policy for Your Classroom

How to write an AI classroom policy

Overview: Write one even if your district has not. Silence is not neutral, and your students are already using these tools in ways you did not intend.

Keep it to one paragraph they will actually read, say what is allowed at each stage of an assignment, and give the reason rather than just the rule.

Most districts still don't have an official AI policy, which means most teachers are deciding this on their own, whether they meant to or not. Silence isn't neutral. Students fill that silence with their own guess about what's allowed, and those guesses are usually more permissive than what you actually intended.

Related reading: How Teachers Can Use AI, our running list of what these tools are actually good for in a classroom, sorted by subject and grade.

Write your own if the district hasn't

You don't need to wait for a district-wide policy to have one. A short, clear statement on your syllabus, stated once at the start of the year and repeated when it's relevant, does more good than no policy at all. Waiting for official guidance just means more months of students operating on assumptions instead of your actual expectations.

Keep it to one paragraph

A policy that runs half a page won't get read. The ones that work are short: what's allowed, what isn't, and what happens if the line gets crossed. Save the nuance for conversation, not the document. A policy's job is to set a clear default, not to cover every possible edge case in writing.

Decide what's allowed at each stage

The clearest policies separate the writing process into stages instead of making one blanket rule. Brainstorming and outlining is a very different use than generating a final draft, and treating them the same tends to either ban something genuinely useful or permit something you didn't mean to. Naming the specific stages where AI is fine, and the ones where it isn't, gives students an actual decision rule instead of a vague sense of what might get them in trouble.

Related reading: Claude for Teachers

Explain the why, not just the rule

Students follow a rule they understand far more consistently than one they were just handed. A policy that explains what you're actually trying to protect, their own thinking, their own writing voice, their ability to do the work without a tool when the tool isn't available, lands differently than a rule that just says no. It also makes the harder conversations easier later, because you're not enforcing an arbitrary line, you're pointing back to something you already explained.

A policy won't prevent every gray area. It will give you and your students a shared starting point instead of each guessing at what the other expects.

Teachers are comparing notes on this in the Faculty Lounge Forums.

Three sample policies you can copy

These are grounded in published district and state policy, not invented. Each one is written to be read aloud to that age group. Sources are listed at the end.

AI Rules for an Elementary School Classroom

In our class, the teacher is the only one who types to the Artificial Intelligence. Students will not have their own accounts, because these tools are made for people 13 and older. When we use it, we will use it together on the big screen, and you will tell me what to ask. AI can sound confident and still be wrong, so we always check. We never use AI to do work.

How to introduce it. Read it standing at the board with the tool already open, so "on the big screen" points at something real. Kids this age accept the account rule easily if you give the reason once and move on. The line that needs a beat is "sound sure and still be wrong." Follow it immediately with a live demonstration: ask the tool something you know it will fumble, let them catch it, and let that be the memory they keep.

Grounded in Bend-La Pine Schools in Oregon, which bars generative AI entirely for K-2 and permits grades 3 to 5 only with teacher supervision inside monitored rooms.

AI Rules for a Middle School Classroom

In our class, you may only use Artificial Intelligence when the teacher says an assignment allows it, and only with the school account. When you use AI, you must include one sentence at the end of the assignment saying what tool you used and what you used it for. That is not a penalty. It is how we keep track of your thinking. And know this: AI writes wrong answers in the same confident voice it writes right ones. It will not always say it's guessing. So you check it, every time. The work is yours to defend. If you are found to have used AI on an assignment that did not allow it, we will have a conversation, and there will be consequences that will likely include you redoing the assignment.

How to introduce it. The sentence doing the most work is "That is not a penalty." Middle schoolers read any disclosure requirement as a trap until you say otherwise, and once they believe you, they disclose freely. Handle the account rule matter-of-factly. If some of your students have not turned 13 yet, that is not a side case at this grade, it is half of sixth grade in September.

Grounded in North Carolina DPI, which splits the middle grades between learning about AI and learning with it at 13 and older, and publishes disclosure examples.

AI Rules for a High School Classroom

On every assignment I will tell you where AI is allowed, stage by stage. Sometimes I will allow it for brainstorming but not for drafting. Sometimes it is closed for every stage of the assignment. The default is that you are not allowed to use it unless I say otherwise.

When you do use AI on assignments I have approved, add one line at the end of the assignment that includes which tool, which stage, what you asked it to do. For example, “I used Gemini to pressure-test my counterargument.”

This policy exists because I can't give you real feedback if I don't know which parts are yours. And you will be asked to account for your own thinking in college and at work. In life, you must be able to explain out loud anything you have produced.

How to introduce it. Do not read this without a concrete stage map in hand for the assignment on the table, or it lands as abstraction. Name the stages out loud: research open, outline open, drafting closed, revision open for grammar only. The "explain out loud" clause is the enforcement mechanism worth leaning on, because it is the same standard a college interview or a first job applies. If college applications are anywhere on their horizon, say directly that admissions essays run on a tighter rule than your class does.

Grounded in Knox County Schools' acceptable use scale for grades 9 to 12 and Washington OSPI's disclosure guidance.

Before you adapt these

Check your district's policy first. Several states now require districts to have one, and yours may already answer questions you were about to answer yourself.

Watch the age line. Most consumer AI tools set 13 as a floor and expect parental consent through 17, so with younger students, run the tool yourself rather than sending them to it.

And do not build any of this on detection software. Seven widely used detectors falsely flagged 61 percent of essays written by non-native English speakers as AI-generated, while performing near-perfectly on native-speaker writing. Vanderbilt disabled its detector and published the math: at a claimed 1 percent false positive rate across the 75,000 papers it processed in a year, around 750 student papers could have been wrongly flagged. Michigan State and Johns Hopkins did the same, and OpenAI withdrew its own detector for low accuracy.

There is strong evidence for bias against English learners and reasonable concern about neurodivergent writers, though the published research on the second is thinner than on the first. Either way, a policy that depends on catching people is a policy that will eventually accuse the wrong student.

If you are writing this for a whole school

Everything above is classroom scale. If you are an administrator setting policy for a building or a district, the questions are different: which tools are approved, who vets the vendor terms, what happens when student data lands somewhere it should not, and who is accountable when an adult lets a tool make a call about a child.

We built a template for that. It covers purpose, scope, permitted and prohibited use, academic integrity, privacy, accessibility, who owns what, training, incidents, and review. The fill-in-the-blank parts are marked, and there are notes throughout on where schools usually get this wrong.

Open the school AI policy template, free to copy, adapt, and republish under your own school's name. No attribution required, no email needed.

One piece of advice before you use it. Most schools do not need a standalone AI policy. You almost certainly already have an acceptable use policy, an academic integrity policy, and a student data privacy policy, and folding these sections into those three is cleaner than creating a fourth document nobody reads. The template is laid out so you can lift sections individually.