If your children use Duolingo, Khan Academy, or Reading Eggs, they’re already using AI. The adaptive learning that makes those platforms useful — the way they adjust difficulty, repeat what isn’t sticking, move on when it is — that’s AI doing what it does well: responding to the individual in front of it, customising the interaction to the data it’s fed.
What’s changed is that the tools are now conversational. You can talk to them, brief them, push back on them. Which makes them more powerful, more useful, and worth thinking about deliberately. AI is now also incredibly attractive for teenagers, meaning you need to be one step ahead to set safety & usage rules. That means understanding what AI is actually useful for and where it can be a less-than-useful learning tool without guidelines in place.
We use AI in three distinct ways in our home-ed. Each is different enough that they’re worth separating out.
As a research tool — carefully
AI is not a good first stop for research, and we don’t use it as one. The problem isn’t just that it gets things wrong (though it does, sometimes confidently). It’s that it does the thinking part for you. When a learner searches for something in a book or via a search engine, their brain has to engage with the question first — what am I actually looking for? With AI, that step disappears. The answer arrives before the question has been properly formed. Done habitually, that’s not a shortcut. It’s a thinking skill quietly not being developed, and ironically, as AI usage becomes more normalised, employers are increasingly asking for proof of critical thinking skills.
So the order in our house is: books first, then human-written internet, then AI. AI comes last, as a check or a supplement, not a starting point.
When we do use AI for research, we mostly use Perplexity. Its default mode provides citations for everything it returns, which means learners can fact-check as they go — the same discipline as book-based research, applied to AI output. It also functions more like an amplified search engine than a generative tool, which keeps the source material traceable. As a side effect, watching Perplexity work is a passive lesson in how citations function, which pays off later in essay writing.
As a tutor
This is where AI earns its place most clearly in our home-ed, and it took us a while to use it this way deliberately rather than accidentally.
AI is a patient, available, infinitely re-explainable tutor. For subjects like maths — where a concept can need approaching six different ways before it’s understood — that’s genuinely useful. We now have an AI model open during teacher-led maths sessions. When my explanation isn’t working, I call in the support teacher. We ask for step-by-step breakdowns, alternative approaches, plain-English explanations of why an answer sheet is doing what it’s doing. Quadratic equations, mostly.
The parameters matter. The AI is there to explain, not to answer. “Show me how this type of problem works” is a different request from “solve this for me,” and that distinction is worth making explicit with your learner before you start.
For writing, Grammarly sits in a similar space: useful as a learning tool if the learner marks their own work first and then uses it to check, not if it replaces the self-editing step entirely. Track changes on; engagement required.
Nb. One way to ensure these rules are being followed through by your homeschoolers is to make them write first drafts of any research-based work and answer maths questions by hand. Even if they’ve breached the book→human-written internet→AI rule, it’s highly unlikely they’ll try to copy-paste in handwriting, and the gaps will be obvious.
As a creative tool
Lower stakes, more open, and genuinely enjoyable. Expressing themselves digitally is an extension of creativity offline; and AI is, in this context, simply another instrument/medium.
Music generation is the one we’ve got most mileage from. SunoAI is our current favourite — you can generate music in any genre, with any instrumentation, from a text prompt. What started as experimentation has quietly become a working knowledge of musical vocabulary and genre: you can’t prompt for “a melancholic baroque piece with harpsichord” without learning what those words mean and whether the output matches them. That’s a music lesson that doesn’t feel like one.
AI image generation is useful paired with other subjects — art, poetry, history. The most productive use isn’t generating finished pieces but using the output as a starting point for discussion: does the image match the prompt? Where did the AI interpret it differently than you expected? How would you tweak it? That conversation is doing more educational work than the image itself.
A note on speech-to-text, which doesn’t fit neatly into any of these three categories but is worth mentioning: for learners who find writing physically or cognitively difficult, speech recognition tools that transcribe spoken words are genuinely useful. Dictating a response, then retyping or copying it out, separates the composition skill from the transcription skill in a way that can unlock things for struggling writers. It’s also useful for you, as the educator, for lesson planning while doing something else entirely.
AI isn’t going anywhere, and treating it as either a threat or a magic solution misses what it actually is: a tool with specific strengths, specific weaknesses, and a significant dependence on the quality of thinking the person using it brings to it.
If you want your home-learners to have a competitive CV, to be able to research in a time-efficient way, and to continue developing self-led learning skills, leaning into AI is important: using it deliberately, talking about it openly and in-depth (ethics, laws and politics included: a longer conversation to be covered in a different post) and treat it the same way you’d treat any other powerful resource – with frameworks and guidance – is the way to do that. The AI rules we use can be found here.
If you want a structured way to teach AI skills, our homeschoolers are currently testing the KTS AI Skills Curriculum, a (free) six-month programme designed for home educators. Sign up to hear when it’s ready.
