Teaching Students to Use AI: Norms, Literacy, and Academic Integrity
Deliberate Academy Editorial Team
Reviewed for accuracy and professional relevance
You're 5 lessons in — don't lose your progress.
Sign up free to save where you are and earn a verified certificate when you pass.
The two most common institutional responses to student AI use are a blanket ban and an anything-goes policy. Neither works. A ban leaves students unequipped for a professional environment where AI fluency is already an expectation in many fields. An unrestricted policy removes the conditions in which genuine learning is required. What works is a third position: AI as a tool that must be used transparently, purposefully, and in ways that preserve the learning that education is actually for.
This lesson builds the framework for that third position — practically, with the specific policies and classroom norms that make it real.
- Distinguish between AI-assisted work and AI-generated work, and why the distinction matters
- Establish clear classroom norms for student AI use without banning tools students will use anyway
- Teach the show-your-thinking principle that preserves learning outcomes
- Set age-appropriate expectations for AI literacy across different stages of education
The Framing Problem
Every educator who has thought seriously about student AI use has encountered the same tension. If students can use AI to produce a written essay in minutes, what does an essay assignment assess? The answer is: it depends entirely on how the assignment is structured and what the student is required to demonstrate.
The framing problem is that most existing assignments were designed in a world where producing a written response required the student to think through the content themselves. That assumption has changed. A student who submits an AI-generated essay has not demonstrated understanding, argument construction, or the ability to synthesize evidence. Or they have — if the assignment requires them to direct, revise, and defend the work in a way that cannot be outsourced to the AI.
The right frame is not "how do we stop AI use?" but "how do we design conditions where learning still has to occur even when AI is available?" That reframe is professionally demanding and worth the effort, because it leads to genuinely better assessment design rather than a technology arms race with detection tools.
The Declaration Principle
Academic and scientific communities already have an established norm for managing undisclosed influence: researchers declare funding sources, reviewers declare conflicts of interest, authors declare contributions. The principle is that transparency, not prohibition, is the mechanism for maintaining trust.
The same principle applied to student AI use shifts the conversation from punitive detection to educational transparency. Instead of "did you use AI?" — which invites concealment — the question becomes "how did you use AI, and what did you contribute yourself?" This is a fundamentally different conversation, and it is a more honest one.
In practice, the declaration principle means building an AI usage statement into assignments. When a student completes an essay, they include a short statement — 50 to 150 words — that specifies which tools they used, what tasks those tools performed, and what work they did themselves. This is not a confession: it is a professional practice, the same kind students will be expected to exercise in many workplaces. Introducing it in education normalizes it as a professional norm rather than an admission of wrongdoing.
Frame the AI declaration requirement as professional practice, not as surveillance. Tell students that researchers, journalists, and professionals in AI-adjacent fields routinely document how AI contributed to their work. The declaration habit they build in your class is preparation for a professional norm they will encounter throughout their careers, not a mechanism for catching them out.
The Show Your Thinking Rule
The deepest problem with AI-generated work is not that it exists but that it is indistinguishable from student work unless the student is required to demonstrate the thinking behind it. An AI can produce a draft. An AI cannot demonstrate that the student understood the source material, made considered editorial choices, or can defend the argument in a conversation.
The show-your-thinking rule builds that requirement into assessment design. It does not prohibit AI use; it requires students to demonstrate understanding in ways that AI use alone cannot satisfy.
This looks different at different levels. A high school student who used AI to generate a first draft might be required to annotate the final version, marking which paragraphs they rewrote substantially and why, which arguments they chose to keep or cut, and what the AI got wrong about their specific source material. A university student might sit a brief verbal examination on a piece of submitted written work — not as a punitive measure, but as a standard part of assessment where the student demonstrates they can discuss, extend, and critique what they submitted.
The verbal component is particularly resistant to AI substitution. A student who directed an AI to produce an essay but did not engage with the content cannot hold a coherent conversation about that essay's arguments. A student who used AI as a drafting tool while genuinely engaging with the material can. The verbal element is not about catching dishonesty; it is about assessing the learning that should have occurred.
Assessment designs that rely primarily on AI detection tools are fighting the wrong battle. Detection tools produce false positives — students accused of AI use when they wrote the work themselves — and false negatives — AI-generated work that passes detection undetected. The more fundamental solution is to design assessments that cannot be completed without demonstrating genuine understanding, which renders detection secondary rather than the primary defense against academic dishonesty.
Introducing the Declaration and Show-Your-Thinking Norm With a Year 12 Cohort
Context
An A-level History teacher at a sixth form college was aware that AI writing tools were widely used by her Year 12 cohort for essay preparation. The college had a blanket prohibition policy, but the teacher recognized it was unenforceable and that it left students without any professional framework for using AI in a way that preserved their own learning. She also knew that some students were using AI as a genuine research and drafting partner while others were submitting unrevised AI output, and the written submissions did not distinguish between the two.
Action
Before the second internal assessment of the year, she held a single 20-minute class discussion introducing the declaration principle and the show-your-thinking norm. She reframed AI use not as a rule violation to be hidden but as a professional practice to be disclosed and managed — and told students that the norm she was introducing was the same kind of transparency expected in academic publishing and professional research. The revised brief required students to include a 100-word AI usage statement, and to be prepared for a five-minute discussion of their essay in their next lesson.
Outcome
Declaration rates in the first submission under the new norm were higher than the teacher had expected: a substantial majority of students disclosed some form of AI use, ranging from research prompting to first-draft generation. The five-minute discussions that followed revealed meaningful differences between students who had engaged deeply with the material and those who had submitted AI output with minimal revision. The teacher described the conversations as the most useful formative assessment she had conducted all year. The norm also removed the adversarial dynamic around AI: students who declared their AI use felt professionally competent rather than caught.
A Year 10 student submits an essay with a declaration stating they used ChatGPT to generate the first draft and then rewrote it themselves. The teacher cannot tell from the written submission how much rewriting actually occurred. According to the principles in this lesson, what is the most educationally sound next step?
Select one answer.
Age-Appropriate Expectations
The appropriate relationship between students and AI tools varies significantly across developmental stages. Applying university-level AI norms to primary school students would be as misguided as treating university students as if they cannot be trusted with these tools at all.
At primary school level, the appropriate framing is AI as a vocabulary helper, a spell-checker, or a question-answering tool for factual queries — similar to how we approach a dictionary or an encyclopaedia. The emphasis should be on understanding that AI answers should be checked, not assumed to be correct, and that writing their own sentences matters for developing literacy.
At middle school level, AI can be introduced as a research brainstorming tool, with explicit teaching about verification. A student who asks an AI for a list of possible reasons for a historical event and then checks each suggestion against their textbook is doing more sophisticated research than one who copies the AI's answer. The skill being developed is not AI avoidance but AI scrutiny.
At high school level, AI as a drafting partner becomes appropriate, with required revision and a reflection statement as part of the submission. Students should be learning to direct AI effectively — which requires knowing enough about the topic to judge the output — and to revise AI-generated text to reflect their own voice and argument.
At university level, AI use as a research and writing partner with full declaration is professionally realistic and educationally appropriate. The skills being developed include directing AI tools effectively, identifying their limitations, and integrating AI-assisted work into a final product that the student can fully defend.
Academic Integrity Reframed
The traditional framing of academic integrity focuses on the product: the essay, the assignment, the submitted work. The question is whether the product was produced honestly. AI complicates this framing because the product can look entirely honest while having been produced without any of the learning that the product was supposed to represent.
A more durable framing focuses on the process: the learning that should have occurred. Academic integrity under this framing asks not "was this work produced without improper assistance?" but "did the learning occur that this assessment was designed to assess?"
Under this reframing, an AI-generated essay submitted as entirely original work is an integrity violation because no learning occurred and the student is claiming credit for learning they have not demonstrated. An AI-assisted essay where the student directed the process, revised every paragraph, can discuss the argument, and declared the AI use may represent genuine learning — and may not be an integrity violation at all, depending on the assignment's design and the institution's policy.
This distinction matters because it focuses educator attention on the right question. The question is not "how much AI was used?" but "how much learning occurred and can the student demonstrate it?"
Classroom AI Norms That Work
Effective classroom AI norms are specific rather than aspirational. "Use AI responsibly" is not a norm; it is a platitude. A norm is a specific, enforceable, educationally justified expectation that students understand before they begin an assignment.
A norm that works might look like this: "You may use AI to generate a first outline for this essay. You must rewrite every paragraph in your own words. You must include a 100-word AI usage statement at the end explaining what you asked the AI to produce and what you changed. Your final submission must be defensible in a five-minute conversation about your argument."
That norm tells students exactly what is permitted, what is required, and what the standard for the work is. It is not punitive, but it makes clear that AI use does not substitute for the student's own engagement. A student who follows that norm has used AI in a way that is educationally legitimate. A student who submits the AI's first outline unchanged has not followed the norm and cannot meet the conversational standard.
Share your classroom AI norms with colleagues. The most disorienting experience for students is when different teachers have completely inconsistent expectations — one bans AI entirely, one allows everything, one has nuanced requirements. Departmental or school-level alignment on a baseline norm removes that inconsistency and helps students develop a coherent understanding of professional AI use, rather than a set of tactical responses to different teachers' individual preferences.
A student submits an essay and discloses they used ChatGPT to write a first draft, then rewrote it themselves. According to the show-your-thinking principle, what additional step would best confirm genuine learning occurred?
Select one answer.
Exercise
Your Task
Write a classroom AI norm for one upcoming assignment — a specific, enforceable statement that tells students exactly what AI use is permitted, what they must do themselves, and what the standard is for the work. Include the AI declaration requirement and at least one show-your-thinking condition. Then write two sentences explaining to students why this norm exists — using the professional framing from this lesson, not a rule-enforcement framing.
Success looks like
- The norm specifies what AI use is permitted — not just 'use it responsibly' but a specific permitted scope such as drafting, outlining, or generating examples
- The norm includes a declaration requirement with a minimum word count and specifies what the declaration must contain
- The show-your-thinking condition cannot be satisfied by submitting unrevised AI output — it requires the student to demonstrate editorial judgment or subject understanding
Watch out for
- Writing a norm that functions as a prohibition disguised as guidance — if every form of AI use is implicitly prohibited, the norm is dishonest about its intent and students will recognize that
- Omitting the 'why' — students who understand the educational reasoning behind a norm are significantly more likely to follow it than students who receive a rule without rationale
Hint
The most effective norms are written for a specific assignment, not for AI in general. 'For this essay, you may use AI to...' is more actionable than a general AI policy statement. Draft it as if you are writing the assignment brief, not a code of conduct.
- The productive frame for student AI use is not prohibition or unrestricted permission but transparent, purposeful use. Ban AI and students are unprepared for professional reality. Allow everything and no learning is required. The third position — AI with declaration, direction, and demonstrated understanding — is the educationally sound one.
- The declaration principle shifts the classroom culture from punitive detection to professional transparency. Instead of asking whether AI was used, the question becomes how it was used and what the student contributed themselves. This models a professional norm students will encounter throughout their careers.
- The show-your-thinking rule builds requirements into assessment design that AI use alone cannot satisfy: annotating editorial choices, revising substantively, and being able to defend work in conversation. These requirements preserve the learning that assignments are designed to produce.
- Age-appropriate AI expectations range from vocabulary helper and spell-checker at primary level to full research and writing partner with declaration at university level. Applying university norms to younger students, or treating university students as if they cannot be trusted with AI tools, are both category errors.
- Effective classroom AI norms are specific: what is permitted, what is required, and what standard the work must meet. Aspirational norms provide no practical guidance and cannot be consistently applied.