Is My Kid Cheating with AI? A Nairobi Parent's Guide to Academic Integrity

How to tell the difference between cheating and learning, what to do when the essay looks too polished, and the one question that settles it.

The essay is unusually polished. The vocabulary has jumped a level. Or you walked past and saw them typing write my conclusion. The reaction is immediate and uncomfortable, and it usually arrives mixed with a second worry: that you do not understand this well enough to handle it.

Here is the distinction that schools rarely explain to parents. Using AI to produce an answer your child cannot explain is cheating. Using AI to challenge, extend or critique thinking they already did is learning. The line is not the tool. It is whether the thinking happened, and whether it happened first.

Copying from the internet left evidence. A parent or teacher could paste a sentence into a search box and find the source. AI-generated text has no source to find, which is why detection tools are unreliable and why several schools that adopted them early have quietly stopped.

So the check has moved from the document to the child. That sounds harder and is actually easier, because you do not need any technical skill to do it.

They cannot explain a sentence from their own submission. Pick one, point at it, and ask what it means. Not a memory test, a comprehension test.

The work appeared with no wreckage. Real work leaves drafts, crossings-out, abandoned openings, notes. Work that arrives complete and clean in twenty minutes usually did not involve much thinking.

They panic when asked to redo one paragraph without the tool. Not annoyance, which is normal for a teenager. Panic, which suggests they do not believe they can.

There are notes or a plan from before anything was opened. Even a scrappy one.

They can tell you where they disagreed with the AI and why. This is the strongest signal there is. A child who overruled the tool was clearly thinking alongside it.

It sounds like them. AI-assisted work that has been genuinely worked on keeps the child's voice, with their odd phrasing and their particular examples. Work that has not sounds like a press release.

Do not open with an accusation. If you are wrong you have damaged the trust you need for every future conversation, and if you are right the child's first instinct will be to conceal rather than explain.

Ask them to talk you through it instead. Genuinely curious, out loud, one section at a time. It becomes clear very quickly, and it is a conversation rather than a trial.

If it turns out they did hand in something they cannot explain, the useful consequence is not a ban. It is rewriting it so they can. That takes an evening, teaches the actual lesson, and is far more effective than removing a tool they will use again in a week.

Most households have never stated a rule, which means the child is inventing one under deadline pressure at 10pm. Say it plainly and in advance.

The version we suggest: you may use AI to help you understand something, to challenge your argument, to check your spelling and to suggest a structure. You may not use it to produce sentences you would not have written and cannot explain. If you used it, say so.

That last clause matters more than it looks. A household where disclosure is normal gets to have the conversation. One where use is forbidden gets concealment and learns nothing.

Many schools have not published a position, and where they have, parents often have not seen it. It is a fair and unremarkable question to put to a class teacher: what is the school's expectation on AI use for assignments, and how is it communicated to students?

Ask it without accusation. Most teachers are working this out at the same time you are, and a parent who asks constructively is more useful to them than one who complains later.

Whatever else you do, use this: can you explain this to me in your own words, without the screen?

It works on any subject at any age, it does not require you to understand the technology, and it cannot be gamed. A child who can explain the work has learned. A child who cannot has handed in something that is not theirs.

This is the same test we use in every AkiliNest session. Children form their idea first, use AI to stretch it, and then present work they can defend with the screen turned off. The habit transfers straight to homework, which is usually why parents notice a difference at home before they notice it in the projects.

Your child asks a voice assistant instead of wondering. Copies ChatGPT without reading. Stops arguing with you because the AI said so. 56% of parents believe AI will mostly harm their teen's critical thinking skills. 51% say it will harm writing skills. Even 66% of teens themselves agree that students might not learn critical skills because AI does the work.

The fear is rational. Cognitive offloading is real: when a tool always provides the answer, the brain stops building the pathway to get there independently.

Three things protect thinking. Think-first protocols, where the child writes or speaks their idea before any tool opens. Presentation requirements, where they explain and defend reasoning aloud rather than just submitting a document. And productive struggle, where adults resist the urge to solve it for them when they are close to a breakthrough.

That last one is the hardest, because watching a child struggle is uncomfortable and rescuing them is fast. The discomfort is the mechanism.

Stage one is absolute trust: 'Mom, Google says.' Stage two is independent judgment: 'Mom, I disagree with the AI because.' The distance between those two sentences is the whole point of AI education.

47% of parents are very concerned about their children getting inaccurate information from AI. The fix is not banning AI. It is teaching children to treat every AI output as a hypothesis to test rather than an answer to copy.

You can start this week without any programme. Ask your child to find one thing an AI told them that they think is wrong, and explain why. The first time they can do it, something has changed.

The AkiliNest bootcamps are built around the think-first cycle described here, and the programme stages show how it changes between ages 8 and 17.

If AI and schoolwork is the specific worry, talk to us about which stage fits your child.