"Public-interest AI" is one of those phrases that sounds like it belongs in a policy paper and nowhere near your family's evening homework routine. But the idea behind it — who a tool is actually built to serve — is one of the most practical questions a parent can ask before letting an AI product anywhere near their child's learning.

Where the phrase comes from

UNESCO's argument at Digital Learning Week 2026 — reflected in the ministerial statement that more than 25 education ministers and representatives adopted on 9 September 2026 — was that education should "remain a common good in the age of AI." Public-interest AI is the practical expression of that: AI systems in schools and learning tools whose design and governance genuinely prioritise a child's learning and wellbeing, rather than treating that as a happy side effect of some other goal — usually engagement, data, or subscription growth.

It's worth sitting with why UNESCO felt the need to say this at all. Nobody in education technology markets themselves as "not in the public interest." The concept exists because good intentions and good incentives don't automatically line up, and a product can be built by well-meaning people while still being optimised — by its metrics, its growth targets, its investor expectations — for something other than a child's actual benefit.

The uncomfortable truth about incentives

Most consumer software, including a lot of education software, is built around engagement metrics: time in app, sessions per week, streaks, notifications that pull a user back. Those metrics work brilliantly for a social media feed. They are a much worse fit for learning, where the actual goal is often the opposite of maximum engagement — you want a child to understand something and then stop, confident enough to move on, rather than stay glued to the screen chasing a reward loop.

This is the tension public-interest AI is trying to name. A tool can be technically excellent, genuinely well-designed, and still be pointed, gently and invisibly, at a goal that isn't quite your child's.

A concrete test, not just a phrase

Because "public-interest" is easy for any company to claim in a mission statement, it's more useful as a question you ask about behaviour than a label you look for in copy. A few genuinely diagnostic questions:

  • What does the product celebrate? Does it congratulate a child for finishing, resting, or understanding something well enough to not need help next time — or does it primarily reward logging in again, extending a streak, or spending more time?
  • Does it have a natural stopping point? Learning tools built for a child's benefit tend to have a sense of "enough for today." Tools built purely for engagement tend not to.
  • Is it honest about its limits? A public-interest-minded product tends to say plainly what it can't replace — a teacher's judgement, a parent's care, real rest. A purely growth-minded one tends to imply it can do everything.
  • Who does the data serve? Is what the tool learns about your child used to help your child, transparently, or fed into something else you can't see or control?

None of these require reading a policy document. They're things you can actually observe in a week of using a product.

Why this matters more, not less, once AI is involved

Before AI, the incentive mismatch between "engaging" and "genuinely helpful" already existed in educational software, but a human — a teacher, a curriculum designer — was usually still deciding what content a child saw next. AI changes that: the system itself is now making moment-to-moment decisions about what to show a child, how to respond, when to nudge them back. That makes the underlying incentive baked into the system matter enormously more, because there's no human in the loop catching the moments where "more engaging" and "genuinely better for this child" pull apart.

That's the real substance behind UNESCO's public-interest framing: it's not an abstract virtue signal, it's a specific worry about who — or what — is actually deciding what happens next in your child's learning, and what that decision-maker has been optimised to want.

What to do with this as a parent

You don't need to become a policy expert to apply this. Next time you're evaluating an AI tool for your child — whether it's something the school uses or something you're considering at home — ask the four questions above, and watch what actually happens over a week of real use, not just what the marketing page promises. The gap between the two tells you almost everything.

FAQ

What does 'public-interest AI' mean in education?

It means AI systems designed and governed with a child's learning and wellbeing as the actual goal — not simply as a byproduct of maximising engagement, data collection, or subscription revenue. UNESCO used the phrase at Digital Learning Week 2026 to argue that education should "remain a common good" even as AI tools enter it.

Isn't every AI education company going to claim they're 'public-interest'?

Yes — which is exactly why the phrase needs a concrete test, not just a marketing claim. Ask what the product actually optimises for: time-on-app and daily engagement streaks point one way; a child's independent understanding and a clear stopping point point the other.

How can a parent tell if a product is genuinely public-interest, not just calling itself that?

Look at what it measures success by. A public-interest-minded tool tends to celebrate a child finishing, resting, or understanding something without help next time — not just logging more minutes. It's also usually honest about what it can't do, rather than promising to replace a teacher or a parent.

Duke Harewood built aitutors.me to measure success the way this article argues it should be measured — by whether a child understood something and can rest, not by how long they stayed logged in. Published 11 September 2026.