Teachomatic what to automate, and what not to

Age Bands: 8, 13 and 17

Start with the honest position, because it is the most useful thing on this page.

There is very little direct evidence about how AI use in learning affects children differently by age. The guidance that exists is mostly about digital media in general. The American Academy of Pediatrics published an updated policy statement in January 2026, Digital Ecosystems, Children, and Adolescents, which treats AI tools as part of a wider digital environment rather than as a separate subject, and which moves away from hour limits toward asking about quality, context, and what the screen time is displacing.

For a workplace-oriented comparison outside education, this page offers another way to look at measurement, workload, behaviour, or accountability.

Some school systems are building differentiated guidance across grade bands on that basis — New York City's public schools have said they are drawing on the 2026 AAP research to do exactly this — but that is inference from general media guidance, applied by educators, not findings about AI in learning.

So treat everything below as reasoning, not evidence. It is what follows from what we know about development and about how the tools fail. Where somebody offers you a confident age-specific rule with a number attached, ask where it came from.

For an external perspective on student digital literacy, privacy, and learning, see Internet Matters.

Reviewed August 9, 2026.

What actually changes with age

Four things, and they change at different rates.

The ability to detect that something is wrong. A younger child has less to check against. An eight-year-old cannot notice that a confidently stated fact is false in a domain they are still learning. This is the strongest argument for restricting unsupervised use with younger children, and it is not about maturity or trust — it is about having a knowledge base to compare against.

Susceptibility to the confident manner. Fluent, authoritative prose reads as true, and the capacity to discount presentation and evaluate substance develops late. Adults are bad at it too.

Where the learning cost falls. The younger the student, the more of what they are doing is foundational practice that cannot be skipped later. Handing over the construction of sentences at eight is different from handing it over at seventeen, because at eight the construction is what is being learned.

What the work is for. By sixteen or seventeen much of the work is preparation for external assessment with its own rules, and the conversation shifts naturally from "what is good for your learning" to "what is permitted, and what you will need to be able to do alone."

Roughly eight

Default: supervised, and mostly not.

At this age most tasks are the foundational practice — decoding, handwriting, arithmetic fluency, constructing a sentence. There is no version of delegating those that is not simply skipping them.

What does work is adult-mediated use: you use the tool, in front of them, out loud. Generating a story to read together, producing five more practice questions, illustrating something. The child sees an adult using a tool and questioning it, which is the actual lesson at this age.

The one direct teaching point that lands: it makes things up, and it sounds exactly the same when it does. Demonstrate once with something they know is wrong. Children find this funny, and funny is memorable.

Independent use is a policy question for your school and their family, and the AAP's framing is the right one — the question is not hours, it is what the activity is displacing.

Roughly thirteen

The age where it actually matters.

They can read fluently, they have a device, and their friends are using it. Almost all the trouble starts here, and so does most of the opportunity.

Three shifts:

Teach it explicitly now, not later. The full lesson is designed for this age upwards. If you wait until sixteen, they will have had three years of unstructured habit formation.

Weight everything toward verification. Concrete, checkable failures — invented sources, wrong dates, confident nonsense in a topic they know. The method matters more here than any rule, because this is where they start using it on work that counts.

Make the cost argument concrete. Not abstract warnings about learning. This specific task is the one the exam tests; using it here will save you an hour and cost you the practice. Thirteen-year-olds respond to specific trades and ignore general ones.

This is also the age where a written class policy starts doing real work, because the ambiguity is genuinely confusing rather than convenient.

Roughly seventeen

Treat them as heading into a world with rules.

They are close to contexts — universities, workplaces, professional bodies — where the norms are set by somebody else and enforced formally. The framing shifts accordingly.

Disclosure as professional practice. Saying what you used and for what is a norm they are about to be held to. Practising it now is preparation, not surveillance.

Judgement over facts. They can check a citation. What they need is the harder thing: recognising the fluent median answer and arguing with the reasoning rather than the facts.

Their own decision about the trade. At seventeen they can genuinely evaluate what they lose by delegating a piece of work. Make the argument once, properly, and then let them make the call on work that is not assessed. Paternalism at this age produces concealment.

What to be careful about

Do not import screen-time numbers into this. Guidance about hours of media use is not guidance about using a tool for twenty minutes on an essay, and quoting it that way is a misuse of the source.

Do not assume older means safer. Sixteen- and seventeen-year-olds are under the most assessment pressure and have the strongest incentive to take the shortcut. Capability is not the binding constraint at that age; stakes are.

Do not assume younger means more vulnerable to accusation. The opposite, often. Older students write more, submit more, and get flagged more — and second-language students of any age are the ones the detectors misjudge.

The short version