Academic Integrity & AI policy template
Integrity policy for the AI era, built on cognitive integrity rather than plagiarism rules: an assist-versus-bypass principle, a four-level assignment designation scale, and assessment guidance that doesn't rely on unreliable detectors.
A starting point, not legal advice — review your adapted policy with district counsel before adoption.
1. Purpose and principle
Academic integrity at [District Name] means that assessed work reflects the student's own learning. AI tools change how that principle is applied, not whether it applies. This policy defines when AI use is appropriate, how it is disclosed, and how the district protects the cognitive work that learning requires.
Why this matters: Layer 3 — Instructional Integration. Anchor the policy in learning, not catching cheaters. The phrase "assessed work reflects the student's own learning" gives teachers a principle that still works when the next tool arrives — a plagiarism-style rule written for essays does not.
2. The core question: assistance or bypass
AI use on schoolwork is evaluated by one question: does it assist the student's learning, or bypass it? AI that helps a student understand, practice, organize, or get feedback generally assists. AI that performs the cognitive work the assignment exists to develop — the reasoning, the writing, the problem-solving being assessed — bypasses learning and is not permitted, regardless of the tool used.
Why this matters: Layer 3 — the cognitive core, and this template's differentiator. Most integrity policies enumerate banned behaviors and are obsolete in a semester. The assist-versus-bypass test gives every teacher and student a durable way to reason about cases the policy never anticipated. It also gives your district a defensible answer to "which cognitive work must remain human."
3. Assignment-level designations
Teachers designate the permitted AI level for each significant assignment, using the district scale: Level 0 — No AI: the work develops or assesses skills that must be demonstrated unassisted. Level 1 — AI for preparation only: brainstorming, studying, and feedback on practice work; none of the submitted product may be AI-generated. Level 2 — AI-assisted with disclosure: AI may support the permitted stages; the student documents what was used and how. Level 3 — AI-integrated: working effectively with AI is part of the assessed skill; disclosure still required. If no level is stated, Level [0/1] applies by default.
Why this matters: Layer 3 + Layer 2. The scale moves the decision to the only place it can be made well — the assignment — while keeping the vocabulary consistent district-wide. Publish the default level so silence is never ambiguous. Grade bands typically default differently (elementary Level 0; secondary Level 1), which keeps expectations age-appropriate without 40 pages of rules.
4. Disclosure
Where a designation permits AI use, students disclose it briefly and honestly: which tool, for what step, and what the student changed. Disclosure is a learning habit, not a confession — accurate disclosure of permitted use is never penalized. Undisclosed AI use where disclosure is required, or AI use above the designated level, is an integrity violation handled under [existing code of conduct].
Why this matters: Layer 3 — Instructional Integration. "Accurate disclosure is never penalized" is the sentence that determines whether students tell the truth. If disclosure feels like self-incrimination, you will get silence and detection-tool arms races instead of the metacognition disclosure is meant to build.
5. Grade-band expectations
Elementary [K–5]: AI use is teacher-directed and whole-class; individual student accounts on generative AI tools are not used. Middle [6–8]: guided individual use of approved tools on Level 1–2 assignments, with explicit instruction in disclosure. High school [9–12]: independent use of approved tools within assignment designations; students are expected to manage disclosure and to build an accurate picture of their own unassisted capabilities before graduation.
Why this matters: Layer 2 + Layer 3. Differentiated boundaries by grade band is one of the first things auditors — and parents — look for, and one K–12 rule is the most common gap in real district policies. Adjust the brackets to your structure; keep the principle that autonomy grows with age.
6. Assessment design
Because detection tools are unreliable and produce false accusations, the district's primary integrity strategy is assessment design, not detection. High-stakes assessments are reviewed against current AI capabilities and use, where appropriate: in-class or supervised components, oral defense or process interviews, drafts and process artifacts, and personalized or local contexts that generic AI output cannot satisfy. AI-detection software output alone is never sufficient evidence for an integrity finding.
Why this matters: Layer 3 — Instructional Integration. Two commitments here protect the district: reviewing high-stakes assessments against what AI can actually do, and refusing to discipline on detector output alone — false-positive rates make detector-only findings indefensible, and they fall hardest on multilingual learners. This section is where the policy earns teacher trust.
7. Responses to violations
Responses follow the district's existing progressive discipline framework, with the learning-centered addition that a first violation ordinarily results in redoing the work under conditions that demonstrate the intended learning. Repeat or egregious violations escalate under [code of conduct]. Records distinguish between AI-related integrity violations and other categories so the district can monitor patterns and adjust instruction and PD.
Why this matters: Layer 4 — Operational Implementation. "Redo it under conditions that demonstrate the learning" keeps the consequence aligned with the purpose (the learning happens after all), and the record-keeping clause gives your governance loop real data about where the policy is straining.
8. Review cycle
This policy and the designation scale are reviewed [annually] by [owner/committee], with input from teachers and students, and immediately after any semester in which AI capabilities materially change what the district's common assessments measure.
Why this matters: Layer 5 — Future-Readiness. Note the trigger is about assessments, not tools: the question to re-ask each cycle is "did the ground under our assessments move," which is the mechanism most likely to catch the next capability shift early.
Integrity boundaries are one layer of readiness
The free 5-Layer Readiness Snapshot scores your district on this and the four other layers of AI governance — ten minutes, personalized report, three priorities.