Academic integrity: using AI in schoolwork

What you'll learn

• What academic integrity is actually protecting, and why that explains the rules
• The spectrum from clearly legitimate AI use to clear malpractice, and where the line sits
• Why disclosure matters, and how to declare AI use properly
• Why "I rewrote it in my own words" does not always solve the problem
• Why AI detectors are unreliable in both directions, and what protects you better
• Why citing a chatbot as a source is usually a mistake
• The self-interested argument: what heavy AI use on coursework costs you in a closed-book exam

Key terms and definitions

Core concepts

What the rules are protecting

Assessment exists to find out what you can do. A mark is a claim about your capability, and it is relied on by teachers, exam boards, colleges and employers.

Once you see that, the rules stop looking arbitrary. If work you submit was produced by something else, the mark describes that thing rather than you — and everyone downstream is misled, including you, because you now have a grade that does not match what you can actually do.

This also explains why the same action can be fine or not depending on the task. Using a calculator is legitimate in most maths assessment and prohibited in a non-calculator paper. Nothing about the calculator changed; what is being measured did.

So the question to ask is never "is AI allowed?" but: what is this task trying to measure, and does using AI here misrepresent that?

The spectrum

Most uses fall clearly on one side. The difficulty is in the middle, which is why it helps to see the whole range.

Clearly legitimate — using AI to learn

• Asking for a concept to be explained differently when a textbook has not worked
• Generating extra practice questions to test yourself
• Asking for the weaknesses in an argument you have written
• Checking your understanding by explaining it and asking what you got wrong
• Asking for feedback on structure or clarity in your own draft

Usually fine, but check and disclose

• Asking for an outline you then research and write yourself
• Using it to find terminology or suggest what to read next
• Having grammar or spelling corrected in work you wrote
• Translating a source you then engage with in your own words

Clear malpractice

• Submitting AI-generated text as your own writing
• Having it produce the analysis or argument you are being assessed on
• Generating references you have not read
• Using it in an assessment where it is prohibited

The reliable test: would you be comfortable telling your teacher exactly what you did? Discomfort is usually the recognition that the mark would no longer describe you.

Disclosure

Where AI use is permitted, say what you did. Not a vague admission — a specific one:

I used an AI assistant to suggest a structure for this essay and to check my grammar. The research, argument and writing are my own.

Good disclosure is specific about the task, honest about extent, and separates what the tool did from what you did. It protects you twice: it is the honest account, and it is evidence of good faith if questions arise later.

Not disclosing when the rules require it is itself a breach, independently of how much help you had.

"I rewrote it in my own words"

Rewriting generated text does not reliably fix the problem, and understanding why matters.

If the ideas, structure and argument came from the model and you changed the wording, the thing being assessed — your analysis — still is not yours. Paraphrasing addresses the surface and leaves the substance.

There is a genuine distinction, though, and it is worth being precise about:

• You asked for an explanation, understood it, and wrote your own analysis → the thinking is yours
• You asked for the analysis, then reworded it → the thinking is not

The difference is not how much text survives. It is whether you could now defend the argument, extend it, or answer a question about it without the tool.

Plagiarism and AI use are related but not identical

Plagiarism wrongs two parties: the original author, who is uncredited, and the reader, who is misled about whose work it is.

With AI-generated text there is usually no specific author being wronged — so the second part carries the whole weight. The offence is misrepresenting your own capability, and it is why "but I'm not stealing from anyone" misses the point.

AI detectors fail in both directions

Schools use software claiming to detect AI-generated writing. You should understand its limits, because they affect you whichever way you use AI.

• False positives are real. Careful, well-structured, grammatically clean writing is what these tools often flag — so strong writers, and students writing in a second language, get flagged for writing well.
• False negatives are easy. Lightly edited text frequently passes.
• They cannot show their reasoning. A score is produced with no evidence you can examine or contest.

Two consequences follow. A detector result is weak evidence and should never be the sole basis for an accusation. And because you cannot rely on being believed on your word alone, the sensible protection is process evidence: keep your drafts, your notes, your version history. A document whose history shows an argument developing over days is far stronger evidence of your own work than any detector score is against you.

Citing a chatbot as a source

Students sometimes list a chatbot in a bibliography. This is usually wrong, for a reason worth understanding.

A source is something a reader can go to and check. A chatbot conversation is not reproducible — the same prompt may give a different answer — and it is not an authority on anything; it produced likely text. Citing it tells your reader nothing they can verify.

If a model told you something you want to use, the right move is to find the real source and cite that. If no real source exists, you have discovered that the claim should not be in your work.

Describing AI use in a methods or acknowledgements note is different and often appropriate. That is disclosure, not citation.

The self-interested argument

Set the rules aside for a moment. Suppose nobody would ever find out.

If AI produced your coursework, you did not practise the thing the coursework was for. Then you sit a closed-book exam, with no tool, on the same material. Coursework is often the preparation for exactly that.

The same logic applies after school. A grade that overstates what you can do places you on a course or in a job where the gap becomes visible, and that is a worse outcome than a lower, accurate mark.

Worked examples

Example 1: Assessing a scenario (4 marks)

A student asks an AI for an essay outline, researches each section themselves, writes the essay in their own words, and does not mention the AI. Assess this.

• The research, argument and writing are the student's own, so the substance is theirs (1 mark)
• Taking a structure from a tool is modest help, comparable to a planning discussion with a teacher (1 mark)
• However, if the rules require disclosure of AI use, not mentioning it is itself a breach (1 mark)
• A one-line declaration would resolve it, so the omission creates a risk for no benefit (1 mark)

Example 2: Explaining why paraphrasing fails (3 marks)

A student generates an analysis, rewrites every sentence, and argues the work is now theirs. Explain why this may still be malpractice.

• The ideas, structure and argument came from the tool; only the wording is the student's (1 mark)
• What the assessment measures is the analysis, which is the part that is not theirs (1 mark)
• A useful test is whether they could defend or extend the argument without the tool (1 mark)

Example 3: Judging an accusation (4 marks)

A student is accused of using AI because a detector gave their essay a high score. Evaluate the strength of this evidence.

• Detectors produce false positives, often on clear, well-structured writing (1 mark)
• They cannot show their reasoning, so there is nothing the student can examine or contest (1 mark)
• A score alone is weak evidence and should not be the sole basis for an accusation (1 mark)
• Drafts and version history showing the work developing would be much stronger evidence either way (1 mark)

Common mistakes and how to avoid them

• Asking "is AI allowed?" Ask what the task is measuring and whether using AI misrepresents that.
• Assuming paraphrasing solves it. If the thinking is not yours, rewording does not make it so.
• Thinking it is fine because nobody is being stolen from. The misrepresentation is of your own capability.
• Treating a detector as proof. It is weak evidence in both directions.
• Relying on being believed. Keep drafts and version history.
• Citing a chatbot as a source. Find the real source, or drop the claim.
• Disclosing vaguely. Say which task, and what was yours.
• Assuming the same rules everywhere. They differ by subject, task and board — check the specific assessment.

Using this in practice

Before using AI on anything assessed:

1. What is this task measuring? Writing, analysis, research, understanding?
2. Would using AI here misrepresent that?
3. What do the rules for this assessment actually say? They vary by subject and task.
4. Could I explain exactly what I did to my teacher without discomfort?
5. If permitted, how will I disclose it? Be specific.
6. Am I keeping drafts? Process evidence protects you.
7. Could I defend this without the tool? If not, you have not learned it yet.

Quick revision summary

• Assessment measures what you can do; work that is not yours makes the mark a false claim about you
• Ask what the task is measuring, not whether AI is "allowed"
• Using AI to learn is legitimate; using it to produce what you submit is not
• Disclose specifically where the rules require it — vague admission is not disclosure
• Paraphrasing generated analysis does not make the thinking yours; the test is whether you could defend it unaided
• The offence is misrepresenting your own capability, which is why "I'm not stealing from anyone" misses the point
• Detectors give false positives and false negatives and cannot show their reasoning — keep drafts as your protection
• A chatbot is not a source; find the real one or drop the claim
• Even with no chance of detection, AI-written coursework leaves you unprepared for a closed-book exam

A note on legal advice

Nothing in this topic is legal advice, and it should not be used as a substitute for it.

Laws differ considerably between countries, they change, and how a law applies depends on facts this material cannot know. Where this topic says something is unlawful or an offence, it is describing the general position in many jurisdictions so that you understand why the rules exist — not telling you what the law is where you live.

What this material is for is helping young people use technology, including AI, appropriately, lawfully and safely, and knowing where to go for help. For anything that has actual consequences, ask somebody qualified: a teacher, your school's safeguarding lead, a solicitor or attorney, your exam board, or the police. If something is happening to you now, do not wait for advice before telling a trusted adult.