"Is it cheating if my child uses AI on homework?" is probably the single most common question parents ask about AI tutoring — and it's also the question UNESCO's 2026 Digital Learning Week guidance treats with more nuance than most school policies manage. The short version: the line UNESCO draws isn't about whether AI was touched. It's about what the AI actually did.

Why "did you use AI?" is the wrong question

Most academic-integrity rules were written for a world where the binary that mattered was "did the student write this, or did someone else." AI breaks that binary, because a single AI interaction can sit anywhere on a spectrum from "explained a concept the student then applied themselves" to "wrote the whole essay for them to submit unchanged."

UNESCO's Digital Learning Week 2026 framing — reflected across the ministerial statement's emphasis on transparency, teacher agency, and public-interest AI rather than a specific rulebook — pushes toward a more useful question: not whether AI was used, but what role it played. That distinction matters because it's the one that actually predicts whether a student learned anything.

The five-minute test

Here's a practical version of the distinction you can use at home, without needing a formal policy document. After your child finishes a piece of homework with AI involved, ask them to redo a similar problem — a different equation, a different paragraph topic — alone, five minutes later, without the tool.

  • If they can: the AI likely acted as a thinking partner. It explained something, asked questions, or checked their reasoning, and the understanding transferred.
  • If they can't: the AI did the thinking, and your child copied the output. The homework got submitted, but nothing was learned — which is the actual problem academic-integrity rules exist to prevent, whether or not any specific rule was technically broken.

This test is more honest than asking "did you use AI," because a student who used AI extensively but as a genuine thinking partner passes it easily, while a student who "just checked their answer" against an AI that quietly did all the reasoning for them fails it.

Why the tool's design decides the outcome before the student does

Here's the part most guidance to parents skips: whether AI use lands as help or as cheating is mostly decided by how the tool is built, not by how disciplined your child is in the moment. A tool designed to answer questions fluently and completely will produce ghost-written homework almost by default, because that's what it's optimised to do — a tired 13-year-old facing a hard question and a tool willing to just answer it is not a fair fight, regardless of how well-intentioned the child is.

A tool built around a hint ladder — nudge, then a bigger nudge, then a fully worked similar problem, but never the answer to the problem actually in front of the child — makes the "thinking partner" outcome the default instead of the exception. That's a structural choice, not a matter of the student's willpower, and it's the reason why some tutors deliberately never give answers.

What this means for a family policy, not just a school one

Schools are still working out formal AI-and-integrity policies, and they vary enormously — some ban AI outright, some allow it with disclosure, some haven't decided yet. Whatever the school's rule, a family can run its own version at home, and it's a better one than most institutional policies manage:

  • Distinguish thinking-partner AI from ghost-writer AI, and apply that distinction consistently rather than treating "AI" as one category to allow or forbid.
  • Ask to see the process, not just the output — a quick "talk me through how you got this" costs two minutes and reveals whether understanding actually happened.
  • Choose tools that make the honest path the easy path. A student shouldn't need willpower to resist a shortcut that a well-designed tool simply doesn't offer.

Where this connects to UNESCO's bigger picture

Academic integrity isn't a side issue in UNESCO's 2026 guidance — it's downstream of the same principles covered in the ministerial statement: teacher agency (a teacher needs to be able to trust that a submitted piece of work reflects the student's own reasoning) and learner rights (a student is entitled to actually build the understanding the work is meant to develop, not just produce an artefact that looks like it). Get the design of the tool right, and the integrity question mostly answers itself.

FAQ

Does UNESCO say using AI for homework counts as cheating?

UNESCO's Digital Learning Week 2026 guidance doesn't issue a blanket "AI is cheating" rule — its framing, reflected in the ministerial statement's emphasis on public-interest AI and teacher agency, is closer to: the question is what the AI actually did in the process, not whether AI was touched at all. A tool that helped a student think through a problem sits very differently from one that produced the finished answer for them to copy.

How can a parent tell the difference between AI help and AI cheating?

A useful test is whether your child could redo a similar problem alone, five minutes later, without the tool. If yes, the AI likely helped them understand something. If they'd be stuck again immediately, the tool did the thinking and handed over the output — which is the pattern UNESCO's guidance and most schools' policies treat as the problem, regardless of whether any rule was technically broken.

What should a school's or family's AI-and-integrity policy actually cover?

The most useful policies distinguish between AI as a thinking partner (explaining, questioning, checking) and AI as a ghost-writer (producing finished text or working for a student to submit) — and apply that distinction consistently rather than banning or permitting AI as a single category. That distinction is closer to what UNESCO's emphasis on transparency and human accountability implies than a flat yes/no rule.

Duke Harewood built aitutors.me so the honest path — thinking it through — is also the easy path, not the one that needs willpower to choose. Published 11 September 2026.