Common Mistakes

8 min

AI for school work: where help ends and cheating starts

Published: 15.09.2026·Updated: 15.09.2026

Dimas Prasetyo

Education Analyst, Algonova

AI for school work: where help ends and cheating starts

AI for school work stops being help the moment your child submits something they cannot explain. That is the whole line, and it settles almost every argument at the kitchen table. Below are 4 questions that test any piece of work in under a minute, what to do if a teacher raises it, and the one marks pattern parents should watch each term.

The four question test

Run these on any piece of homework. Two clear noes and the work is not ready to be handed in.

Can your child explain the middle steps, not just the answer?
Did they type the final version themselves, in their own words?
Has every number been checked manually?
Would they get a similar result in a test with no device?

The fourth question matters most in Indonesia, because tests are still handwritten with nothing on the desk. A child who passes questions one to three usually passes the fourth. A child who fails the fourth has a gap, and the homework mark is covering it up.

Help, grey area, cheating

The difference is not the tool, it is which part of the thinking was skipped.

What the child does	Verdict
Asks AI to explain a concept, then solves it alone	Help
Asks it to rewrite a question in simpler language	Help
Asks it to generate extra practice questions	Help
Asks it to check working the child already did	Grey, depends on the teacher
Asks it to outline an essay, then writes it	Grey, better declared
Asks for the answers and copies them	Cheating
Pastes an AI-written essay and changes a few words	Cheating

The grey rows are settled by one thing only: whether the teacher knows. A declared outline is safe in almost every classroom.

An Indonesian student doing homework at a desk with books and a laptop

The marks gap that gives it away

Watch the distance between homework marks and test marks across one term. Everything else is guesswork; this is measurable.

Pattern over a term	What it usually means
Homework high, tests high	The learning is real
Homework high, tests falling	Thinking is being handed to the tool
Homework low, tests fine	Effort problem, not a learning gap
Both drifting down	The topic itself is not understood

Row two is the one to act on, and it is invisible if you only open the exercise book. Two months is usually enough to see it. When that pattern appears, the fix is not a stricter ban but finding which topic the child stopped understanding.

How to raise it with the teacher

Ask the school's rule before you need it, not afterwards. Most schools in Indonesia have no written AI policy yet, so the teacher's own rule is what applies.

"Is my child allowed to use AI to understand material at home?"
"If they use it for an essay outline, should that be declared?"
"Is there anything you want done without AI entirely?"
"How would you prefer parents to check homework at this level?"

Those four questions take two minutes at a parent meeting and remove almost every later misunderstanding. If the teacher has no position yet, the four question test above is a reasonable default, and it is stricter than most policies.

If your child is caught

Separate the two problems before you react. One is honesty, the other is a learning gap, and they need opposite responses.

Ask what the task was and what they actually did, without raising your voice
Ask which part they could not do. That answer is the real information
Agree a redo on the same topic, with you nearby
Apologise to the teacher together, briefly and without long excuses
Change the setup so it is not repeated: same room, history visible, phone elsewhere

Most children do not use AI out of dishonesty. They use it at the exact moment the work stopped making sense, usually late at night with nobody to ask. Punishing the shortcut without fixing that moment simply moves the habit to a friend's phone.

Rebuild the homework routine

Move the hardest part earlier and give it a person, not a chatbot. Four changes cover most cases:

Homework starts before dinner, not after nine when help is unavailable
The child writes down the exact sentence that stopped them, before opening anything
One weekly slot where a teacher answers those stuck points
The device stays in a shared room, so the rule is enforceable without arguments

The stuck-point note is the cheapest of these and the most useful. After two weeks you have a written list of exactly which topics are missing, which most parents never get.

What is reasonable by age

The younger the child, the more the adult does the asking. Independence here comes from demonstrated checking, not from a birthday.

Age	Reasonable	Not yet reasonable
7 to 9	Parent asks AI to explain, child listens	Child having their own account
10 to 12	Child asks with a parent nearby, history visible	Unsupervised use late at night
13 to 15	Own account with permission, history open	Deleting chats, essays written for them
16 to 17	Mostly independent, rules agreed in advance	Undeclared use in graded work

The jump between rows two and three is where most conflicts appear, usually because the rules were never discussed before the account existed. Setting them in the first week costs nothing and prevents the argument entirely.

An Indonesian child writing a solution in a notebook beside a closed laptop

Turn the tool into a skill instead

The children who genuinely benefit are the ones taught to direct AI, check it and know when to ignore it. That is a teachable skill, and it looks nothing like chatting.

Ask for a method, then solve it yourself
Ask it to test you rather than answer for you
Check anything numerical by hand, every time
Spot invented references and say so plainly
Know which tasks it simply handles badly

A child who can do all five is better prepared than one kept away from AI entirely, and far better prepared than one who leans on it. That is what we build in our AI classes for kids, alongside coding.

What to do this week

Run the four question test on the next piece of homework, without announcing it
Compare the last homework marks with the last test marks in one subject
Ask the teacher the four questions above at the next opportunity
Move homework earlier and put the device in a shared room
If a gap appears, take a class on that topic rather than tightening the rules

A gap found this month is a small problem. The same gap found in the final exam is not. Homework marks are a poor early warning; the comparison with tests is a good one.

Worth reading next: ChatGPT for kids, learning AI for primary kids and what parents say.

Frequently asked questions

Is using AI for homework cheating?

It depends which part of the thinking was skipped, not on whether a tool was open. Asking AI to explain a method and then solving the questions yourself is studying. Asking for the answers and copying them is submitting work that is not yours. The practical test that settles almost every case at home: can your child explain the middle steps, did they write the final version themselves, were the numbers checked manually, and would they get a similar result in a test with no device. Two clear noes and the work is not ready to be handed in.

What is the rule in Indonesian schools?

Most schools have no written AI policy yet, so in practice the teacher's own rule is what applies. Until that changes, treat undeclared AI use in graded work as someone else's work, because that is the safest assumption. The useful move is to ask before you need the answer: whether AI may be used to understand material at home, whether an AI-assisted essay outline should be declared, and whether anything must be done without it. Two minutes at a parent meeting removes almost every later misunderstanding.

How do I know whether my child used AI?

Look for three things at once rather than just one. The work is unusually clean with no crossings out, the word choice sounds like an adult wrote it, and it took far less time than usual. Any one of those can be innocent; all three together rarely are. The most reliable check is not detection software, which is inaccurate and produces false accusations. It is a conversation: ask your child to explain the middle steps or to teach you the same topic in five minutes. A child who did the work can do this easily.

My child was caught. What should I do?

Separate honesty from the learning gap, because they need opposite responses. Start by asking what the task was and what they actually did, without raising your voice. Then ask which part they could not do, because that answer is the real information. Agree a redo on the same topic with you nearby, apologise to the teacher together briefly, and change the setup so it is not repeated. Most children use AI not out of dishonesty but at the moment the work stopped making sense.

Will AI make my child weak at maths?

It can, but only if it replaces the working rather than the explanation. The risk is concrete in Indonesia, because tests are handwritten with no tool on the desk, so a homework habit that skips the thinking builds a gap that eventually appears on the exam paper. The measurable warning sign is the distance between homework marks and test marks over a term. Homework high while tests fall is the pattern to act on, and it is invisible if you only open the exercise book.

Should AI be banned completely at home?

A ban on output fails because it cannot be enforced and simply moves the behaviour to a friend's phone. Rules about process work, because they change what the tool is for. The ones that hold up in practice: ask it to explain rather than answer, write the final version yourself, check every number manually, submit nothing you cannot explain, keep chat history visible for the first months, and never enter personal details. Families who set these early argue about it far less.

Can AI be used to check my child's answers?

This sits in the grey area and depends entirely on the teacher, so it is better declared. Checking has a real benefit: a child who solves first and then verifies is doing exactly what good students do. The problem is that AI is unreliable at arithmetic and tends to agree when challenged, so it can confirm a wrong answer with full confidence. If your child uses it to check, the rule must be clear: the tool shows how to verify rather than simply saying right or wrong.

What about essays and compositions?

The wording has to be your child's, with no exception worth making. An AI-written essay with a few words changed is still an AI-written essay, and teachers recognise the register quickly. Using it to find angles or produce an outline is defensible in most classrooms, provided it is declared. The reason for the hard line here is not only honesty. Writing is the skill being assessed, so handing the sentences to a machine removes the entire purpose of the task, unlike handing over a maths explanation.

At what age can a child use AI independently?

Independence comes from demonstrated checking rather than from age, though the pattern is fairly clear. From 7 to 9, the parent asks and the child listens. From 10 to 12, the child asks with a parent nearby and history stays visible. From 13, an own account with permission makes sense, since 13 is also the minimum the main services set, with history open to you. From 16, mostly independent with rules agreed in advance. Conflicts most often appear at the jump to an own account.

How do I turn this into something useful?

Teach the five things that actually transfer: ask for a method rather than an answer, have the tool test the child, check every number manually, spot invented references, and know which tasks it handles badly. A child who can do all of these is better prepared than one kept away from AI entirely and far better prepared than one who leans on it. That is what our AI classes cover alongside coding. The free trial class shows within an hour whether your child suits this way of working.

Blog

More from Blog

Algonova worldwide

Algonova Mexico Algonova Malaysia Algonova Global

Contact

Whatsapp: +62 21 50996971

Email: info@algonova.id

Address

Coworking Area, Gedung IT 6 Lantai Dasar, Nongsa Digital Park, Jalan Hang Lekiu, Kelurahan Sambau, Kecamatan Nongsa, Batam, Indonesia Directorate General of Consumer Protection and Trade Compliance – Ministry of Trade of the Republic of Indonesia. WhatsApp Ditjen PKTN: 0853-1111-1010.

AI for school work: where help ends and cheating starts

The four question test

Help, grey area, cheating

The marks gap that gives it away

How to raise it with the teacher

If your child is caught

Rebuild the homework routine

What is reasonable by age

Turn the tool into a skill instead

What to do this week

Frequently asked questions

More from Blog

Online vs In-Person Coding Classes for Kids: Which Fits Better? (2026)

When Should Primary School Kids Start Coding: A Readiness Guide

8 Best Kids Coding Schools in Indonesia (2026): Review & Comparison

The marks gap that gives it away

How to raise it with the teacher

If your child is caught

Rebuild the homework routine

What is reasonable by age

Turn the tool into a skill instead

What to do this week

Frequently asked questions

More from Blog

Online vs In-Person Coding Classes for Kids: Which Fits Better? (2026)

When Should Primary School Kids Start Coding: A Readiness Guide

8 Best Kids Coding Schools in Indonesia (2026): Review & Comparison