Start with one answer your child has already received. Ask them to identify a claim, find an independent source and explain what they would change if the evidence disagreed. Confidence with an app is useful; being able to question its output is a different skill. The family examples below are editorial illustrations using pseudonyms, not accounts of particular students.
The Question Behind The Finished Homework
A polished paragraph can hide an unfinished understanding. Imagine Leena, a parent looking at a beautifully organised science explanation written by her thirteen-year-old, Maya. Instead of guessing whether AI was involved, Leena asks Maya to explain one sentence without reading it. In this illustrative example, the next conversation concerns the idea itself: which part does Maya understand, which part needs checking, and what help would allow her to explain it independently? That is a more useful starting point than treating every smooth sentence as evidence of cheating. Parents can ask about the process before judging the product. The same question works when a child has used a video, a search engine or a friend's notes.
What The Recent Research Actually Says
Ofcom's UK research summary, published on 21 May 2026, reports AI use among 56% of eight- to seventeen-year-olds, rising to 66% among sixteen- and seventeen-year-olds. It also identifies limited understanding of AI and uncertainty about checking its accuracy. These findings describe a research population; they do not tell you what your own child knows. UNICEF Innocenti's September 2026 report draws on consultations with 97 children aged nine to seventeen in the Democratic Republic of the Congo, Egypt, Mozambique, Senegal and Rwanda. Participants discussed verification, privacy, subject knowledge and independence. This qualitative work offers perspectives, not a representative survey of UK children. Both sources make it reasonable to discuss understanding alongside access.
Choose A Claim Small Enough To Check
Set aside ten minutes and choose a harmless factual answer: a description of a historical event, an explanation of a science concept or a summary of a book you have both read. Avoid turning the first exercise into a test of private worries or a consequential decision. Ask your child to underline one checkable claim, rather than deciding whether the whole answer feels convincing. Now separate the claim from its presentation. A neat table, precise date or reassuring tone does not supply evidence. Write the claim on paper in ordinary language. If it becomes difficult to explain without the surrounding paragraph, begin by working out what the words mean.
Leave The Chat To Find The Evidence
Checking should involve opening an independent source, not simply asking the same chatbot whether it is certain. Choose an appropriate reference: the original organisation for its own announcement, a museum or archive for a historical document, or the relevant course provider for its course information. Open the page and find the supporting passage. A link that exists can still fail to support the particular claim. Keep three brief notes: what the AI said, what the source says, and what remains unresolved. Sometimes the source confirms the answer. Sometimes it narrows it. Sometimes no adequate evidence turns up. Being able to leave a question open is part of the exercise, rather than an embarrassing failure to finish.
Notice Dates, Places And Missing Conditions
Illustrative student Arun asks for a shortlist of degree courses and receives a convincing comparison. Before using it, he and a parent select one entry to examine. Which university is it? Which campus? Which course year? Does the page describe the qualification Arun actually studies? Is a suggested subject essential, recommended or merely useful? This example does not establish eligibility or recommend a course. It shows why a plausible answer can become misleading when conditions disappear. Younger children can try a simpler version using a local activity: confirm the venue, date and age group. Older teenagers can practise finding exclusions and checking whether a page has been replaced. The difficulty should fit the child's understanding.
Keep A Piece Of Thinking That Belongs To Them
Agree on a small task your child will do before requesting help. It could be writing an initial explanation, sketching a solution or listing questions about a passage. After using a permitted tool, ask them to mark what changed and why. This creates something concrete to discuss when the final answer looks dramatically better than the first attempt. For schoolwork, check the teacher's instructions before using AI; different tasks may permit different kinds of assistance. If the rules are unclear, ask the school rather than treating this guide as permission. A useful family habit is to distinguish support with understanding from submitting work the child cannot explain. The aim is visible learning, not a competition to produce the most impressive document.
Make Personal Information A Separate Decision
An illustrative younger child, Theo, wants a personalised revision plan and begins adding details about classmates, teachers and family routines. An adult can help him describe the task without supplying those identities. A fictional timetable and broad learning goal may be enough for the exercise. Information about someone else should not be treated as his to share simply because it makes the prompt more detailed. Before using a service, adults should check its age requirements and privacy information and follow school guidance where relevant. Do not assume every product has the same settings or permissions. A family discussion can begin with three categories: information we can omit, information we need to ask about, and information we will keep out of the tool.
Keep Human Conversations Available
Not every request needs a technical answer. A young person asking a tool what to do after an argument may need someone who knows the situation and can listen. Parents can make that route easier by describing when and how their child can approach them, another trusted adult or a school contact. Avoid promising that any automated response understands the child simply because it sounds sympathetic. You do not need to inspect every conversation to teach this distinction. Discuss hypothetical examples together: help organising a timetable, disagreement with a friend, a confusing school instruction. Ask what the tool could reasonably contribute and what a person should help resolve. Keep the invitation to talk calm and specific.
Try A Family Check Without A Score
Take turns bringing one ordinary claim to the table. The adult should go first and explain a mistake or an uncertainty of their own. Each person identifies the claim, selects a source and describes the result. Give credit for a clear explanation of the check, including a conclusion that the evidence is insufficient. Do not award points for sounding certain. Repeat the exercise occasionally with different formats: a paragraph, an image caption or a proposed course comparison. If images are involved, avoid relying only on visible oddities such as strange hands; ask where the image came from and what evidence connects it to the event claimed. Keep the activity short enough that checking feels usable outside the exercise.
Your Next Step
Tonight, ask your child to show you one example of how they use AI, if they are comfortable doing so. Choose a non-sensitive task and work through one claim together. Write down a single change to your household routine: perhaps opening the source before repeating a fact, attempting homework first, or asking an adult before sharing personal details. Return to that agreement after a week. Ask whether it helped and where it was awkward. You are building a habit of explaining, checking and choosing help thoughtfully. You do not need to settle every argument about AI at once. A child who can describe why an answer needs checking has something more useful than a perfectly polished paragraph.
Sources And Further Checks
Primary sources checked on 06-10-2026. Course and policy details can change; check the current official information before acting.
Ofcom: UK Children's Media Lives And AI Use, 21 May 2026 UNICEF Innocenti: Skills For An AI World, September 2026AI-assisted production. How We Produce Stories · Report A Correction
