A one-page family sheet sorts chatbot help from the homework itself

When a child asks, “Can I use a chatbot for homework?”, “yes, but use it responsibly” sounds reasonable and barely helps with the next question. May they ask what a word means? May they ask for an outline? If the chatbot writes a complete paragraph and the child changes a few words, does that still count as their own work?

A more useful rule has to be concrete enough that the household can look at one situation and classify it the same way. This article helps a parent make a one-page rule sheet for chatbot use during homework. The aim is not to ban AI, and it is not to find a way around the school’s rules. The aim is to keep the part the student is supposed to show in the student’s hands, while still allowing AI to help in the places the family and the teacher accept.

On 30 June 2026, UNICEF published an analysis of new data from 10 countries estimating that at least 20 million children have used AI, and that about 13 million said they use AI to support learning and homework. The UNICEF release of 30 June 2026 says children in this 10-country data are taking up AI faster than adults — more than three times faster — and it does not measure a knowledge gap between parents and children. These figures are not statistics for Vietnam alone, but they show that “when may my child ask a chatbot?” is already a real household question, not only a forecast.

UNICEF Vietnam currently has a Vietnamese page, Parenting in the AI era, that encourages parents to explore AI with their children, talk about what a tool can and cannot do, and look together at which parts of an answer are useful or suspicious. UNESCO also stresses a human-centred approach, data protection, and age-appropriate use of AI in education.

The example below is a situation and output constructed by the editors. It is not the result of a test on a particular AI model, account, or school.

First, ask what the assignment is trying to check

Suppose the child has a constructed assignment:

Read a 300-word passage about planting trees in the city. Write about 120 words explaining two benefits of trees, and use at least two details from the passage as evidence.

Before opening a chatbot, the parent and child underline what the teacher may be looking for: the child can choose two benefits, find evidence, and explain in their own words how the evidence connects to their point.

If the chatbot does all three and the child only copies or swaps a few words, the final product can look polished while the parent no longer knows whether the child can do the thinking the assignment is asking for.

That is the root rule of the sheet: AI may help around the learning goal; do not let it quietly replace the goal itself. If you are not sure what the assignment is trying to check, asking the teacher is better than guessing.

Three zones, not a ban list

Take one sheet of paper and make three columns.

Zone	When to use it	Example with the assignment above
May ask	AI helps with a tool, a word, or a process, and does not produce the part that must be submitted	Explain “green space” in language for a 7th-grade student. Do not write the assignment paragraph for me.
Do it first, then ask	The child needs to produce a first answer so a parent can still see the real skill	The child chooses two benefits and two pieces of evidence, then asks: Point to any sentence in my paragraph that does not clearly explain the link to the evidence. Do not rewrite it for me.
Do not give this to a chatbot / stop	AI would do the part the teacher needs to assess, or the request needs data that should not be shared	Write a complete 120-word paragraph for me to submit.; a photo with the child’s full name, class, grades, or account

These three zones deliberately avoid a good/bad label. The same AI question can fit practice homework and not fit a test or a class that bans AI. The teacher’s and the school’s rules always come first.

Try the sheet on three concrete questions

Still using the constructed assignment, the child is about to send three prompts.

Prompt A: What are urban trees? Explain in 3 easy sentences.

If the class materials allow looking up a concept, this can sit in May ask. After reading, the child should close the chatbot and say it again in their own words. If they cannot, the AI answer has not become their knowledge.

Prompt B: Here are the two points I chose and two evidence sentences. Ask me about each one so I explain why the evidence supports the point. Do not write the paragraph.

This belongs in Do it first, then ask. The chatbot plays the person who asks back, and the explanation still comes from the child. A parent can check by looking at the draft before the AI question: did the two points and the evidence already exist?

Prompt C: Write the best 120-word paragraph from this reading so I can submit it.

If the goal is to assess the student’s ability to choose evidence and write an explanation, this crosses the line. Put it in Do not give this to a chatbot. The problem is not that an AI paragraph is certainly wrong. The problem is that it hides the evidence of what the child can do alone.

This is a deliberate boundary check, not an “AI error” observed in a test. The three prompts above are editorial examples. They are not output that was run on a fixed AI model, and no chatbot is claimed to have produced a specific output.

Before submission, do not only ask “did you use AI?”

A more useful short check asks the child to reproduce the thinking without looking at the chatbot. With the example above, a parent can ask:

1. What two benefits did you choose?
2. Which sentence in the reading is the evidence for each point?
3. Why does that evidence support your point?
4. Is there any sentence in the submission you cannot explain yourself?

If the child can answer and the draft shows that the core came from the child, AI is more likely supporting the learning than hiding it. If the child cannot explain a sentence already written in the work, stop there: remove that sentence, go back to the source, and rewrite it from their own understanding.

This check also avoids an argument that is hard to settle about the percentage that is “AI-written.” What matters at the homework table is which skill the teacher is assessing, and whether the child can actually show that skill.

Add a separate row for the child’s data

A family rule sheet should include one rule that does not depend on the subject:

Do not paste a full name, school or class, address, phone number, account, password, health record, grades tied to an identity, or someone else’s private documents into a chatbot just to finish homework faster.

UNESCO notes that using generative AI usually means the user shares data with a provider, and it recommends protecting personal data and age-appropriate use. UNICEF also stresses that parents need to understand the tool with their child, rather than assuming the child already knows every risk.

If an assignment contains real data about a student or a family, make a simulated or anonymized version if the teacher allows it, or do not use an outside service. For a child’s account, a parent also needs to check the current age terms of that specific service and the rules where they live. UNESCO’s 2023 guidance discusses a minimum age of 13 for independent use in an education context. That is not a statement that every service has the same terms, and it is not a statement that age 13 is the law that applies generally in Vietnam.

A rule sheet can be written in five minutes

You can use the template below and change it to match the school’s rules:

CHATBOT RULES FOR HOMEWORK

1. Before asking AI, the child says what skill this assignment is checking.
2. MAY ASK: a hard word, a concept explanation, extra practice questions — if the teacher allows it.
3. DO IT FIRST: the main point, the evidence, the calculation, the reasoning, or the draft the teacher needs to assess.
4. DO NOT ASK AI TO PRODUCE THE SUBMISSION if that is the part the child must show.
5. Do not put personal or private data into a chatbot.
6. Before submitting, the child must be able to explain every part of the work.
7. If the teacher’s rules differ from this sheet, follow the teacher.
8. If you do not know whether a kind of AI use is allowed, stop and ask an adult or the teacher.

Do not print the sheet and treat it as finished. Try it on one real assignment that is not sensitive this week. If two people classify the same situation differently, the rule is still vague and needs to be rewritten.

When the sheet is not enough

Stop using a chatbot for that assignment if you cannot tell which part the teacher wants the student to do alone, if the work is a test or an assessment that bans AI, if the tool asks for data you do not want to share, or if the child starts depending on answers they cannot explain back.

Do not turn the parent into someone who supervises every click. UNICEF suggests exploring AI with the child and using questions about what the tool does well, what is confusing, and how it produces an answer. A good rule sheet should make that conversation easier: the child knows the boundary before opening a chatbot, and the parent has a way to check the learning without becoming an AI expert.

The final aim is not a piece of homework with “no trace of AI.” The aim is that after using AI, the child still owns the understanding the assignment was given to build and to assess.

Further reading

1. UNICEF Vietnam — Parenting in the AI era: explore a chatbot with a child and ask together which parts of an answer are useful or not yet right. The page is in Vietnamese.
2. UNESCO — guidance on technology and AI in education, available in Vietnamese: a Vietnamese page for reading the guidance; the specific clauses are in the primary document below.

Primary sources

1. UNICEF — press release of 30 June 2026: estimates that at least 20 million children have used AI, about 13 million use it for learning, and children in the 10-country data are taking up AI more than three times faster than adults.
2. UNICEF Innocenti — snapshot of children and parents using AI: the same 20 million and 13 million estimates, from data in 10 countries, not a Vietnam statistic.
3. UNESCO — Guidance for generative AI in education and research, 2023: a human-centred approach, data protection, and a discussion of age 13 as a threshold for independent use in education; age 13 is not the law in Vietnam.