How to write assignment-level AI directions students can actually follow

Quick answer

Effective assignment-level AI directions name the learning goal, list allowed and prohibited actions, identify approved tools and data limits, state disclosure and evidence requirements, explain what the student must be able to do independently, and describe consequences or next steps. A reusable district scale can create consistency, but each assignment still needs a plain-language statement.

Why district policy is not enough

A district policy can define approved tools, privacy rules, and broad academic-integrity expectations. It cannot know why a specific teacher assigned a lab conclusion, vocabulary exercise, coding project, personal narrative, or timed essay.

The appropriate boundary depends on the learning objective. AI feedback may strengthen revision practice in one task and erase the skill being assessed in another. The teacher needs a fast way to translate the objective into a rule students can understand before they begin.

The seven-part assignment statement

A complete direction does not need to be long. It does need to close the most predictable gaps. Students should not have to infer whether brainstorming counts as use, whether grammar tools are included, whether a chatbot may see the source text, or whether disclosure is required.

Use the same seven elements across classes so students learn where to look.

1. Purpose

Name the skill or understanding the assignment is intended to show.

2. Allowed actions

List concrete uses such as asking for explanations, generating practice questions, receiving feedback, or checking syntax.

3. Prohibited actions

List actions that would replace the assessed work, such as generating the final response or rewriting entire sections.

4. Approved tools and data

Identify required school accounts and information that must not be entered.

5. Disclosure

State what tool, prompts, outputs, or description of use must be submitted.

6. Independent responsibility

Explain what the student must verify, understand, defend, or reproduce without AI.

7. Review process

Explain what happens if use is unclear or appears inconsistent with the directions.

Four copy-ready levels

A simple scale gives teachers shared vocabulary while preserving instructional choice. The labels matter less than the behavior definitions. Avoid level names that imply moral judgment; a no-AI assessment and an AI-integrated project can both be legitimate.

These examples should be adapted to grade level, subject, accommodations, and approved district tools.

• Level 1, independent work: Do not use generative AI to plan, draft, solve, revise, or check this assignment. You may use the accommodations and non-generative tools listed by your teacher.
• Level 2, AI tutoring: You may ask the approved AI tool for explanations, examples, or practice on the concept. Do not ask it to produce any part of the submitted response. Submit a brief description of what you asked.
• Level 3, AI feedback: Complete your own draft first. You may request feedback on clarity, reasoning, organization, grammar, or test cases. You must decide which suggestions to use and submit the original draft, revised draft, and disclosure.
• Level 4, AI collaboration: AI use is part of the task. Follow the required workflow, preserve prompts and outputs, verify factual claims and sources, and be prepared to explain every submitted decision.

Write verbs, not vibes

Words such as ethical, appropriate, limited, and responsible are useful principles but poor standalone instructions. Replace them with observable actions: brainstorm five questions, explain a concept, critique an outline, identify a bug, propose test cases, rewrite a sentence, generate a paragraph, or produce the final answer.

Also name edge cases. Spellcheck, translation, autocomplete, citation tools, image generation, code completion, voice transcription, and AI features embedded in familiar software can blur the boundary. Students should not be punished for failing to guess which product feature the teacher had in mind.

Align disclosure with the amount of use

Disclosure should produce useful evidence without turning every assignment into paperwork. A tutoring-only task may need one sentence naming the tool and purpose. A collaborative AI project may need prompts, selected outputs, revisions, source checks, and a reflection.

The disclosure should answer three questions: what did the tool do, what did the student do, and how did the student verify or transform the result? Copying an entire transcript is rarely necessary unless the process itself is being assessed.

Preserve accommodations and equitable access

A no-AI direction should not accidentally prohibit assistive technology, translation, speech-to-text, text-to-speech, or supports required by an IEP or Section 504 plan. Teachers need district guidance for distinguishing generative assistance from approved accommodations and for handling tools that combine both.

If AI use is required, the district must provide access to an approved tool and a meaningful alternative when the workflow is inaccessible or inappropriate. Students should not be expected to buy a subscription, expose personal data, or use a personal account to complete required work.

Teach the direction before enforcing it

Model one or two examples and non-examples. Show students how a permitted tutoring request differs from asking for the final answer. Let them practice writing a disclosure. For younger students, use icons or short labels consistently across the learning-management system and paper assignments.

When a boundary is crossed, first ask whether the direction was clear, visible, and taught. Strong enforcement begins with strong notice.

Frequently asked questions

Should every assignment include an AI statement?

Assignments where AI could materially affect the assessed work should include a clear statement. A consistent level label plus a short task-specific explanation is usually enough.

Is grammar checking considered AI use?

It depends on the tool and assignment. Teachers should state whether basic correction, generative rewriting, translation, and style suggestions are allowed rather than expecting students to infer the boundary.

What should students disclose when they use AI?

At minimum, the tool and purpose. More substantial collaboration may require prompts, selected outputs, drafts, revisions, source verification, and a reflection on what the student accepted or rejected.