National survey data collected in the past year found that fewer than half of principals said their school had any AI policy, and only about a third of teachers reported clear academic integrity guidelines. Meanwhile, adoption has surged, with more than half of teachers using AI for instruction or administrative work. That mismatch leaves individual teachers to decide on their own what is acceptable, which produces inconsistency and anxiety. A short, well-written staff policy is one of the cheapest ways to close that gap.

Many leaders delay writing a policy because they assume it must be exhaustive and legally airtight. In practice, a long framework is rarely read, and it goes stale as tools change. A one-page document that answers the questions teachers actually ask has a better chance of shaping behavior. It can later be expanded into a longer policy once the school has more experience.

Teachers in several recent state and national surveys reported that they had not seen the guidance their governments had already issued. In one national review, nearly three out of four teachers said they had never encountered the ministry guidance on generative AI. Brevity and direct distribution matter as much as content. Put the policy where staff already look, such as the faculty handbook and the beginning-of-year slide deck.

What the one page should say

Start with purpose in two sentences: why the school allows limited AI use by staff and what outcomes it hopes to protect, such as student privacy and fair grading. Follow with scope, naming which roles and which activities the policy covers, including lesson planning, drafting feedback, and communicating with families. Then state the central rule plainly, which in most schools should be that a teacher remains responsible for every grade and every comment a student receives. Everything else on the page supports that principle.

• Approved tools: a named list, with a process for requesting additions.
• Human review: every AI-drafted score or comment is read and approved by a teacher before release.
• Student data: no names, identifiers, or sensitive details entered into tools that are not approved.
• Transparency: teachers tell students and families when AI assists with feedback.
• Contact person: one named individual who answers questions and logs concerns.

Defining human review so it means something

The phrase human in the loop appears in many policies without any definition, which makes it easy to satisfy on paper and ignore in practice. Spell out what review requires, for example that a teacher reads the student's work alongside the draft feedback and confirms that the score matches the rubric. Clarify that clicking approve on a batch without reading does not count. A definition this concrete gives administrators something to audit and gives teachers a standard to meet.

It also helps to describe what review might look like at different volumes. A teacher with 30 essays can read each one carefully, while a teacher with 150 might review every score and spot-check comments in detail. The policy does not need to dictate minutes per paper, but it should make clear that a larger load does not lower the standard. Teachers who understand that distinction will be more candid about how they actually work.

Protecting student data and trust

Data protection is where district counsel and teachers most often talk past each other. Explain in plain language that student work and identifying details should only enter tools the district has reviewed, and that any tool used for grading should be able to describe where data is stored, who can access it, and whether it is used to train models. If a tool cannot answer those questions, it should not be used with student writing. A short checklist for vendors, appended to the page or linked from it, makes this practical.

Trust with families depends on candor. A sentence in the syllabus or class newsletter saying that AI may help draft feedback and that the teacher reviews all results prevents surprises later. Parents rarely object to a transparent workflow, but they react strongly when they discover undisclosed use. Building disclosure into the policy removes that risk and gives teachers a ready answer if a parent asks how feedback on their child's essay was produced.

Rolling it out and reviewing it

Introduce the policy at a staff meeting and invite questions instead of simply emailing it. Share two or three concrete scenarios, such as using AI to draft comments on a set of essays or asking a chatbot to rewrite a student's paragraph, and ask teachers to decide whether each is allowed. The conversation will reveal gray areas you can clarify in the next version. Professional development was reported as ineffective by many teachers in recent surveys, so keep sessions short and practical.

Plan to review the policy every semester, since tools and state rules change quickly. Collect questions in a shared document throughout the term and use them as the agenda for revision. After a year, you will have both a better policy and a staff that understands why each line is there. That shared understanding is the real goal, with the page itself serving as a reference.