AI in school: spotting AI-written work and setting tasks AI can't do for you
In short: you cannot reliably prove a text was written by AI today — detectors make mistakes, and they especially often accuse students writing in a non-native language. So the workable teacher strategy isn't catching, it's redesigning assignments: tie them to personal experience, to your class's own data, and to a short oral defence. Then AI stops being a shortcut and becomes, at most, a draft tool.
Why detectors are a bad foundation
The temptation is obvious: paste the text, see "87% AI", conversation over. The problem is there's no evidence behind that number. Detectors don't "see" a text's origin — they estimate how predictable and smooth it is. And smooth, predictable writing isn't unique to models.
- False positives for non-native writers. A student writing in a second language uses simpler constructions and standard phrasing — exactly the signals a detector reads as "machine-like". This is a known, well-documented flaw of these tools.
- False positives for careful students. Tight structure, neutral tone, no errors — and honest work scores a high "AI percentage".
- Evasion is trivial. Ask the model to write simpler and livelier, rewrite a couple of phrases by hand, and the score drops. It's a filter in reverse: it catches the honest and lets through anyone who tried to hide.
- A percentage isn't evidence. "87%" is not "87% probability" — it's a number on an opaque scale. You cannot present it to a student as proof, methodologically or ethically.
The conclusion is blunt but important: you cannot build an accusation on a detector. At most it's a reason to look more closely and to talk — and that talk starts with a question, not a verdict. More on how these tools work and why their accuracy is overrated: AI text detectors: do they work.
What works instead
Shift the focus: not "how do I catch them" but "how do I make copying pointless". AI is excellent at depersonalised tasks — "write an essay about friendship", "make a report on volcanoes". It's poor at tasks tied to a specific person, a specific moment and specific data that isn't on the internet.
Six ways to redesign an assignment
- Tie it to personal experience. Not "an essay about friendship" but "a moment in your life when you realised friendship isn't always fun — describe the place, the time and what exactly you felt". A model will invent it, and the invention collapses at the first follow-up question.
- Tie it to the lesson. "Use three examples we covered on Thursday and explain why the fourth example in the textbook doesn't fit here." AI doesn't have your Thursday.
- Give them your data. A class survey, lab measurements, a table you built yourself. The student works with AI on top of that data — which is a legitimate skill, not cheating.
- Ask for the process, not the product. A draft, an outline, a list of rejected ideas and why they were dropped. AI writes the final text in a second; it can't produce the history of your doubts.
- Oral defence. Three minutes, two questions about the text: "explain this paragraph in your own words" and "what would you change if…". It's the fairest and fastest way to know whose work it is, and it needs no technology at all.
- Legalise AI and demand reflection. "You may use AI. Attach your prompts, attach what the model produced, and write what you changed and why." Copying turns into analysis — and, honestly, into a better lesson.
A prompt to help you redesign a task
You are a curriculum designer. Help me rework an assignment so it can't be completed with a single AI prompt. Subject: [SUBJECT], age group: [AGE] Current assignment: [TEXT] Learning goal: [WHAT THE STUDENT SHOULD LEARN TO DO] Give 3 reworked versions. For each: - what exactly ties it to the student's personal experience, our class's own data, or a specific lesson - which process artefact they submit (draft, outline, choices) - 2 questions for a 3-minute oral defence - how to grade it (criteria that reveal understanding, not textual smoothness) Don't propose bans or detectors. Preserve the learning goal.
How to know the redesign worked
A simple honesty test: take your new assignment, paste the whole thing into ChatGPT and see what comes out. If the model produces text a student could submit as is, you haven't redesigned the task — you've just made it longer. If the model produces a plausible hollow shell that collapses at "so where did these numbers come from?", you're on the right track.
Second signal: a student who did the work honestly should be able to defend it for three minutes with no preparation. If they can't, either they didn't understand it, or the assignment tests something other than what you intended. Both findings are useful, and neither needs a detector.
How to talk to your class about it
A flat "no AI" fails for the same reason banning calculators failed: the tool is in everyone's pocket, and students use it regardless of your opinion. What works is an agreement stated out loud.
- Name the rules explicitly. Where AI is fine (finding ideas, explaining something unclear, spellcheck), where it isn't (submitting generated text as your own), and what must be disclosed if used.
- Explain why. Not "because it's dishonest" but "because what you submit isn't text, it's your understanding; text without understanding is worthless to me and to you".
- Teach them about hallucinations. Students genuinely don't know that models confidently invent facts, dates and sources. One live demo — ask AI about a book that doesn't exist — beats ten lectures. The mechanics: AI hallucinations.
- Never accuse on a percentage. If a paper raises questions, ask them: "walk me through how you wrote this." A wrong accusation costs far more than a missed case of cheating: a class's trust takes years to rebuild.
Where AI genuinely helps the teacher
The flip side: the routine that eats your evenings automates surprisingly well.
- Generate 15 variants of the same problem type with different numbers
- Rewrite a text to your class's level, or adapt it for a student who's struggling
- Build a grading checklist and oral-defence questions
- Invent a 5-minute warm-up, debate or role-play
- Draft feedback on a paper — which you then rewrite for that specific child
One hard boundary: don't paste student work with names, grades or any personal detail into a chat. That's children's personal data, and rules like the GDPR apply in full. Anonymise before sending, or use tools your school has approved. More: privacy when working with AI.
The bottom line
Catching is pointless, detectors are unreliable, bans don't work. Something else does: assignments where the value is created by a human and AI is only a tool. That's more work up front and far fewer conflicts later. If you want to understand these models yourself first, start with the beginner's guide to ChatGPT and our article on prompts.
FAQ
Can a detector prove a student used AI?
No. Detectors estimate how predictable a text is, not where it came from, and they regularly err in both directions. Their output is not evidence and cannot justify an accusation or a lower grade. At most it's a reason to have a calm conversation about the work.
Why do detectors misfire so often on non-native speakers?
Because they react to simple constructions, standard phrasing and predictable vocabulary — exactly how a person writes in a second language. The tool therefore systematically hits students who are already vulnerable. That's a core reason not to use it as proof.
What if I'm almost certain a paper was written by AI?
Start with a conversation, not an accusation: ask the student to explain a couple of paragraphs in their own words and describe how they got there. A three-minute oral defence tells you more than any percentage. And weigh the cost of error: a false accusation damages trust more than a missed case of cheating.
Which assignment can AI definitely not do for a student?
One that needs data not on the internet: personal experience with specifics, material from your particular lesson, your class survey results, your lab measurements. Plus anything requiring the process — drafts, rejected ideas, an explanation of choices.
Should I ban AI in class?
A blanket ban usually fails: the tool is available to everyone all the time, and you have no way to verify compliance. An explicit agreement works better: where it's allowed, where it isn't, and what must be disclosed. Then you're managing the use rather than pretending it doesn't exist.
How do I grade work when a student says they used AI?
Grade understanding, not smoothness. Ask for the prompts, the raw model output and a description of the edits — then grade the quality of those edits, the choice of sources and the oral defence. It's fairer and teaches exactly the skill they'll need later.
Should I teach students about AI hallucinations?
Absolutely. Many genuinely believe the model 'knows' the answer and doesn't err. Show them live how it confidently invents a fact or a non-existent source — that's a critical-thinking lesson and the best vaccine against thoughtless copying at once.
Can I upload student work to ChatGPT for grading?
Not with names, grades or personal details. That's children's personal data, and responsibility sits with your school and with you, not the service. If you want help grading, anonymise the text or use tools your school has officially approved.
How long does redesigning assignments take?
The first one takes an hour or so; after that it speeds up — the patterns repeat and apply to a whole topic at once. AI helps here too: give it the current assignment and the learning goal and ask for three versions tied to personal experience, with oral-defence questions.