Generative AI Policy for Schools: Balancing Innovation and Academic Integrity

Executive Summary

• Generative AI presents unique challenges for schools because it creates content that closely mimics human-produced work, fundamentally challenging how we assess learning
• Detection tools are unreliable — building policy around catching AI-generated submissions is a losing strategy that also harms innocent students
• Assignment design is your most powerful lever — assessments that resist AI substitution protect integrity better than surveillance
• Academic integrity isn't dead, it's evolving — the skills being assessed and how we assess them need to adapt, not the underlying principle
• Prohibition drives usage underground — students will use these tools; the question is whether they learn to use them responsibly
• This generation needs AI literacy — schools have an obligation to prepare students for AI-augmented workplaces
• Clear categories of AI-permitted use help students and teachers understand expectations
• Process documentation (showing work, explaining reasoning) becomes more important than final products

Why This Matters Now

ChatGPT launched publicly in November 2022. Since then, generative AI has transformed from a novelty to an everyday tool for students worldwide. Your students are using these tools — the only question is how.

The generative AI difference:

Unlike earlier AI tools that analyzed or processed information, generative AI:

• Creates original text, images, and code that can be mistaken for human work
• Produces outputs of sufficient quality to complete many academic assignments
• Improves rapidly, making today's detection approaches obsolete quickly
• Is freely and widely available to students of all ages

The stakes:

Schools face a choice between:

1. Prohibition — Attempting to ban and detect, which fails and misses the educational opportunity
2. Permissiveness — Allowing unrestricted use, which undermines learning objectives
3. Purposeful integration — Defining when and how GenAI use is appropriate, preserving learning while building AI literacy

This post outlines approach #3.

Definitions and Scope

What Is Generative AI?

Generative AI refers to AI systems that create new content based on prompts. In educational contexts, this primarily includes:

What Makes GenAI Policy Different?

Your general AI policy covers AI broadly. A GenAI-specific policy or policy section addresses:

• Academic integrity implications unique to content generation
• Assignment design considerations
• Disclosure requirements for AI-assisted work
• Subject-specific guidance (GenAI in English vs. Maths vs. Art)
• Assessment adaptation strategies

The Detection Problem

Why detection-based policy fails:

1. Accuracy is poor. Current detection tools have false positive rates of 10-30%, meaning innocent students are accused. They also miss AI content, especially when edited.

2. Tools degrade over time. As AI improves, detection becomes harder. Detection tools trained on GPT-3.5 struggle with GPT-4.

3. Gaming is easy. Simple paraphrasing, translation round-trips, or asking AI to write in a different style defeats most detection.

4. False positives harm students. Being accused of cheating when you haven't cheated is traumatic and can have lasting impacts.

5. It's an arms race you'll lose. Resources spent on detection aren't spent on education.

The implication:

Policy cannot depend on catching AI-generated work. Instead, design assessments and processes that make AI use either:

• Irrelevant (the task requires something AI can't do)
• Visible (the process reveals whether AI was used appropriately)
• Permitted (with clear guidelines)

SOP: Designing GenAI-Aware Assessments

Purpose

This procedure guides teachers in designing assessments that maintain academic integrity and learning objectives in the context of readily available generative AI tools.

Assessment Design Principles

Before designing any assessment, consider:

1. What skill or knowledge am I assessing?
2. Can generative AI perform this task? How well?
3. What process or product demonstrates that a student has the skill?
4. What AI use, if any, supports learning rather than replacing it?

Step 1: Classify the Assessment Type

Step 2: Select an Appropriate Strategy

Strategy A: Process-Over-Product

• Require documented drafts, revision history
• In-class writing components
• Oral defense of written work
• Reflection on writing/thinking process

Best for: Essays, research projects, creative writing

Strategy B: AI-Resistant Design

• Hyperlocal topics (our school, our community)
• Personal experience requirements
• Very recent events (post AI knowledge cutoff)
• Integration of live class discussions

Best for: Any assignment where unique context is available

Strategy C: AI-Inclusive Design

• AI use explicitly permitted with disclosure
• Assessment focuses on prompting, evaluation, and improvement of AI output
• Comparative analysis (student work vs. AI work)

Best for: Building AI literacy, teaching critical evaluation

Strategy D: Authenticated Assessment

• In-class, supervised conditions
• Oral examinations
• Live demonstrations
• Practical assessments

Best for: High-stakes assessments, skill verification

Step 3: Document Expectations

For each assessment, communicate clearly:

• AI policy for this specific assignment
• What types of AI use are permitted/prohibited
• Disclosure requirements if AI is used
• How the assessment will be evaluated
• Consequences for policy violations

Step 4: Review and Iterate

After implementation:

• Gather student feedback
• Assess whether learning objectives were met
• Identify issues or gaming
• Refine for next iteration

Generative AI Use Categories

Help students and teachers by establishing clear categories:

Category 1: No AI Permitted

AI tools may not be used at all for this task.

When to use:

• Assessing baseline writing skills
• Foundational skill demonstrations
• Examinations
• Specific learning objectives requiring unaided work

Example: "Write an in-class essay analyzing the themes in Chapter 3. No AI tools may be used."

Category 2: AI for Research and Brainstorming Only

AI may be used for ideation and information gathering, but not for drafting.

When to use:

• Research projects where synthesis is the skill
• Creative projects where originality matters
• Early-stage learning of a skill

Example: "You may use ChatGPT to brainstorm topic ideas and clarify concepts, but all writing must be your own."

Category 3: AI as Editor/Reviewer

AI may be used to improve student-created work.

When to use:

• When communication quality matters alongside content
• Professional writing practice
• Non-native English speakers

Example: "Write your first draft independently. You may then use AI to check grammar and suggest improvements, but the ideas and structure must be yours."

Category 4: Full AI Collaboration

AI may be used throughout, with disclosure.

When to use:

• AI literacy learning objectives
• Professional simulation (AI use expected in field)
• Focus on evaluation and judgment skills

Example: "You may use AI tools freely for this project. Submit your final work along with your prompt history and a reflection on how AI contributed."

Step-by-Step Policy Implementation

Step 1: Develop Policy Framework

Work with academic leadership to establish:

• Default AI use category for assessments
• Subject-specific variations (Art department may differ from English)
• High-stakes assessment protocols
• Examination policies

Timeline: 2-4 weeks

Step 2: Train Teachers

Professional development on:

• Understanding GenAI capabilities and limitations
• Assessment design strategies
• Clear communication of expectations
• Handling suspected violations

Timeline: 1-2 sessions, ongoing support

Step 3: Communicate to Students

Roll out to students through:

• Assembly or class introduction
• Clear documentation in student handbook
• Subject-specific guidance from teachers
• Examples of acceptable vs. unacceptable use

Timeline: 2-4 weeks for initial rollout

Step 4: Communicate to Parents

Inform parents about:

• School's approach to GenAI
• Why this approach was chosen
• How assessments are being adapted
• How to support at home

Timeline: Newsletter + optional information session

Step 5: Implement and Monitor

During implementation:

• Collect teacher feedback on assessment approaches
• Track any integrity concerns
• Gather student feedback
• Monitor parent questions/concerns

Timeline: Ongoing

Step 6: Review and Adapt

Regular review:

• Annual policy review (minimum)
• More frequent review if technology changes significantly
• Incorporate learnings from implementation

Common Failure Modes

1. Detection-Dependent Policy

The problem: Policy that relies on catching AI use creates false accusations and misses actual violations.

The fix: Design assessments that don't depend on detection. Focus on process, oral components, and authenticated work.

2. Blanket Prohibition

The problem: Banning all AI use drives it underground and misses the educational opportunity.

The fix: Create clear categories of permitted use. Teach responsible AI practices.

3. Subject-Blind Policy

The problem: One-size-fits-all policy doesn't account for different subject needs (AI use in coding vs. creative writing differs).

The fix: Allow subject departments to adapt policy within a common framework.

4. Product-Only Assessment

The problem: Only assessing final products makes AI substitution easy.

The fix: Include process elements, drafts, oral defense, and reflection.

5. Unclear Expectations

The problem: Students and teachers unsure what's allowed, leading to inconsistent enforcement.

The fix: Clear categories, assignment-level guidance, and explicit communication.

6. Ignoring AI Literacy

The problem: Treating AI only as a threat misses the opportunity to prepare students.

The fix: Include AI literacy as a learning objective. Teach critical evaluation of AI outputs.

Generative AI Policy Checklist

Policy Development

• GenAI-specific policy or section developed
• AI use categories defined
• Default category established
• Subject-specific variations documented
• Assessment design guidance provided
• Teacher training planned

Communication

• Policy communicated to teachers
• Policy communicated to students
• Policy communicated to parents
• Assignment-level expectations template created

Assessment Adaptation

• Existing assessments reviewed against GenAI
• High-risk assessments adapted
• Process components added where appropriate
• Oral/authenticated options available

Monitoring

• Teacher feedback mechanism in place
• Student feedback mechanism in place
• Review schedule established

Metrics to Track

Tooling Suggestions

Assessment Design Support

• Turnitin AI Writing Detection — Use with caution; supplement, don't depend
• Google Assignments — Process tracking, draft comparison
• OneNote Class Notebook — Process documentation

AI Literacy Teaching

• ChatGPT, Claude — Direct experience with GenAI
• AI comparison exercises — Compare AI to human work
• Prompt engineering projects — Teach effective AI use

Policy Communication

• Learning Management System — Central policy location
• Subject handbooks — Subject-specific guidance

Frequently Asked Questions

Next Steps

Generative AI has fundamentally changed the academic integrity landscape. Schools that adapt thoughtfully can maintain rigorous standards while preparing students for an AI-augmented future.

For guidance on developing your school's GenAI policy and adapting assessments:

Book an AI Readiness Audit — Our education experts help schools navigate the GenAI challenge with practical, workable policies.

Related reading:

• How to Create an AI Policy for Your School: A Complete Guide
• AI Acceptable Use Policy for Schools: Separate Templates for Students and Staff
• ChatGPT Policy for Schools: Specific Guidelines for Students and Teachers