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

Learn Coding

Fajar Hidayat
Coding Curriculum Team, Algonova

Prompt writing for kids is the part of AI that survives the next tool change. Chat windows will look different in three years; asking precisely for a method, a constraint and a check will not. A child aged 9 and up can master the 4-part structure below in one sitting, and it takes about two weeks of small daily use before it becomes a habit.
Role, task, constraint, check. Every prompt worth writing has those four parts, and children memorise them faster than adults do.
| Part | What it does | Example wording |
|---|---|---|
| Role | Sets who is answering | "You are a patient maths teacher" |
| Task | Says what to produce | "Explain how to add fractions" |
| Constraint | Limits the answer | "For a 10 year old, under 100 words" |
| Check | Forces verification | "Then ask me two questions to test I understood" |
The check is almost always skipped, and it is precisely what turns a chatbot into a study partner. Without it, the child just reads an answer. With it, the child has to produce something.
A lazy prompt asks for the answer. A directed prompt asks for the process. Same tool, completely different outcome for the child.
| Lazy | Directed |
|---|---|
| "What is 3/4 plus 2/5?" | "Show me the steps to add 3/4 and 2/5, without the final answer" |
| "Write an essay about recycling" | "Give me three angles for a recycling essay, then critique my opening line" |
| "Is this right?" | "Do not say right or wrong. Show me how to check it myself" |
| "Explain photosynthesis" | "Explain photosynthesis for a 9 year old, then quiz me" |
Show your child the left column once, then the right column, and ask which taught them more. Most children pick correctly without being told.

Start here, in this order. Each one teaches a different move, and together they cover most of what a student needs.
Number seven changes the whole dynamic, because the child has to answer rather than read. Number eight builds scepticism, the habit most worth having.
Younger children need the constraint written for them; older ones write it themselves.
| Age | What they can do | What the adult supplies |
|---|---|---|
| 7 to 8 | State what they want in one sentence | Role, constraint and check |
| 9 to 10 | Use the four parts with a written reminder | Occasional correction |
| 11 to 13 | Write full prompts and refine them | Almost nothing |
| 14 to 17 | Chain prompts, compare outputs, spot invented sources | Nothing |
The jump between 10 and 11 is clearly noticeable. Before it, children accept the first answer. After it, with practice, they start pushing back and asking for a different explanation, which is exactly the behaviour you want.
Reading the answer critically matters more than writing the prompt. A perfect prompt followed by blind acceptance teaches nothing.
The second last point makes a good weekend exercise. Two different answers to the same question, side by side, shows a child that this is a text generator rather than a source of truth. Nothing you say conveys that as quickly.
Directing a tool precisely is the same skill as writing a clear brief, and it shows up in almost every job today. Children who practise it get three things at once:
That is why prompting sits next to coding in our AI classes for kids rather than standing alone. Directing software without understanding how it works usually stalls at the first unexpected result.
Under 13 there is no personal account, so practise the thinking with tools built for children. The structure still transfers even when the interface differs.
The paper version is the most effective and costs nothing. Children discover for themselves that "draw a house" produces ten different houses, and that is the entire lesson about constraints.

If day six of week two works, the skill has stuck. Being able to explain the structure without the card is the honest test, and most children need about two weeks of small daily use.
Prompting is a multiplier, not a substitute. A child who does not yet understand fractions will get a fluent explanation of fractions and still be unable to do the homework, because reading an explanation is not practising. When that gap appears, the answer is a teacher who watches the working, not a better prompt.
Worth reading next: ChatGPT for kids, AI for school work and our coding programme.
Prompt writing is asking an AI tool precisely enough that the result is useful, and it turns out to be the same skill as writing a clear brief for a person. It matters for children because this is the part that survives the next tool change. Chat windows will change in a few years, but stating what you want, setting a constraint and building in a check will still work. There is a second benefit parents usually notice first: children who practise it get better at saying what they actually mean.
Four parts: role, task, constraint and check. Role sets who is answering, for example a patient maths teacher. Task says what to produce, for example explain how to add fractions. Constraint limits the answer, for example for a 10 year old in under 100 words. Check forces verification, for example then ask me two questions to test my understanding. The check is almost always skipped and matters most, because without it the child simply reads an answer.
Around 9 is the comfortable starting point for the full structure, though younger children can practise the thinking without any AI at all. At 7 or 8, the adult supplies the role, the constraint and the check while the child says what they want. At 9 to 10, children use the four parts with a written reminder nearby. From 11, most can write full prompts and refine them after a weak answer. Remember that the main chat services set 13 as the minimum age for a personal account.
Use tools built for children, since the structure transfers even when the interface differs. Teachable Machine has the child define categories, supply examples, then watch the model get confused by careless input, which is constraint and check in physical form. Scratch teaches the same discipline by making a sprite do exactly what it is told. The cheapest exercise needs no screen at all: one child writes instructions for drawing a shape, another follows them literally, and half of them draw the wrong thing.
A lazy prompt asks for the answer, a directed prompt asks for the process. Instead of what is three quarters plus two fifths, the directed version asks for the steps without the final answer. Instead of is this right, it asks the tool to show how to check it. Instead of explain photosynthesis, it asks for an explanation aimed at a nine year old followed by a quiz. The tool is identical in both cases; what differs is whether the child ends up with a finished answer or with something they had to work through.
It helps a child get a better explanation faster, which is real but limited. Prompting is a multiplier, not a substitute. A child who does not yet understand fractions will receive a fluent explanation of fractions and still be unable to finish the homework, because reading an explanation is not practising. That distinction is worth stating clearly at home. If the same gap keeps appearing on the same topic, a better prompt will not close it, and a teacher who watches the working will.
Ask them to teach you the four parts from memory, with no card on the desk. That is the honest test, and it usually takes about two weeks of small daily use. Two more signals are worth watching. First, whether they refine a prompt after a weak answer instead of simply accepting it, which shows they understand the tool responds to how you ask. Second, whether they check numbers manually without being reminded, which shows the scepticism has formed.
No, and teaching that is more valuable than the prompting itself. Three exercises make the point faster than any lecture. Ask the same question twice in separate chats and compare the difference. Give it a correct sum, insist it is wrong, and watch it apologise and switch to a mistaken answer. Ask for a source and then try to find that source together. Children who do all three once stop treating the output as fact.
Not yet, which is why we teach it alongside coding rather than as a separate subject. Directing software without understanding how it works usually stalls at the first unexpected result, because the child has no picture of what went wrong. Coding supplies that picture: variables, sequences, conditions and the habit of fixing errors. A child with both can ask precisely and then judge whether the answer makes sense, and that combination is far more useful than either half alone.
Through projects rather than lists of tips. The child builds something small, reaches a point where the tool gives an unhelpful answer, then learns to restate the request with a clearer constraint. The teacher watches the process rather than the final output, and that is where the real correction happens. Prompting sits alongside coding at every level, so children learn to direct a tool while understanding what the tool is doing. The free trial class shows within an hour whether your child suits this way of working.