Homework & revision help · ages 8–18
An AI your child can use — and you can trust.
Even the new AI “study modes” are just a setting your child can switch off. Bearings is built so it can’t be: it works through the hard bits until they get unstuck, but it won’t hand over the answer — so the thinking stays with them and the learning is real. It explains things through what they love, knows the curriculum they’re actually studying, and lets you see they’re still learning.
Right now, an AI can finish your child’s homework in seconds — and you’d never know whether they learned anything, or just copied it. That’s the fear. This is built so that can’t happen — and so you can finally see the learning is real.
It gets them unstuck — without doing it for them.
Stuck on a problem, your child gets real help: it asks what they’ve tried, points them to the next move, and works through the tricky part with them. Ask it to “just give me the answer” and it won’t — a separate check stops it, every reply, every time.
The thinking has to happen in your child for the work to get done at all.
And you can see they’re still learning.
The hard part of AI and homework is not knowing: are they actually learning, or just letting it think for them? You get a parent’s-eye view — topics worked on, what to revisit, how much help they needed.
Never a score, never ranked against other children. And your child sees exactly what you see. Insight, not surveillance.
It explains things through what they love.
Football, horses, music, Minecraft — tell the tutor, and it reaches for that to make a tricky idea click: the same concept, mapped onto something they already understand. You choose what it knows — it never guesses or builds a profile.
And when a comparison only goes so far, it says where it stops.
It knows the curriculum they’re actually studying.
Built around your child’s own curriculum — American or English — so it coaches toward the work their teacher expects, not a generic version.
It guides them toward what good work looks like. It never marks them, and never writes it for them.
Built differently for every age
An 8-year-old and a 17-year-old need different things. So they get them.
The promise never changes — it won’t hand over the answer. What changes is how it talks, how much it helps, and what you hear as a parent. Pick an age:
How it actually helps
Six ways to move them forward — none of them “here’s the answer”.
Asks the question that gets them to the next step themselves.
The gentlest nudge first, less as they get it — never the answer.
They put the method in their own words and try a fresh example — so you know it stuck.
Asks them to justify their reasoning — without saying right or wrong.
Truly stuck? It works a different problem through what they love, then hands theirs back.
Worried? It stops and points them to a trusted adult or helpline — and still won’t do the homework.
How it compares
The same homework question — three very different answers.
| ChatGPT Study Mode | Gemini Guided Learning | Bearings |
|---|
Study modes are a real step forward — we just go one level deeper. Comparisons reflect the products as publicly documented; we date our claims.
Built around their curriculum
Not generic help — the work their school actually sets.
For families on a US curriculum — built around Common Core, NGSS, AP frameworks and the Readers & Writers Workshop model, so it coaches toward what those classrooms expect.
For families on the English curriculum — the national curriculum and GCSE (AQA, Edexcel, OCR), coaching toward the objectives their exams are actually marked on.
You tell us which curriculum your child follows — it coaches toward what good work looks like, and it never marks your child or writes the work for them.
Its strongest ground — problems with a definite answer it can rigorously hold back.
Coaches structure and reasoning; we’re honest that essays are harder to guard than equations.
Explains the method and stress-tests reasoning across subjects.
The proof, not just the promise
Why you can trust it — built into the product.
- It coaches toward the next step instead of handing over the answer.
- A separate check holds back any reply that would give the answer away.
- Built around your child’s own curriculum — American or English — and it never marks them.
- Personalises through interests your child chooses — never ones it guesses about them.
- No scores, grades or rankings about your child — anywhere, for anyone.
- You set it up and give consent; you can erase everything with one request.
- If something’s worrying your child, it points them to a trusted adult or a helpline.
Why it works — the thinking behind it
Built on how learning actually works — and honest about what the research does and doesn’t say.
Explain a hard idea through something a child loves and they engage more, hold it in mind more easily, and remember it better — a 2024 meta-analysis of 34 studies found medium-to-large effects. The evidence it helps them carry a skill to a brand-new problem is real but lighter — so we say “grounded in”, never “proven to raise grades”.
A good analogy carries across the structure of an idea, not just the surface — long-established in cognitive science. A careless one plants a misconception, so it flags where an analogy holds and where it breaks.
“Study modes” coach by an instruction the same AI can ignore — OpenAI’s own docs say theirs “will provide a direct answer” sometimes. Ours is an independent check the answer has to get past first.
What it will do
- Coach your child to the next step — and get them genuinely unstuck.
- Fade its hints as they get it; explain through interests they choose.
- Coach toward what good work looks like in their curriculum.
- Give you an honest view that they’re learning (rolling out) — shared with your child too.
- Point them to a trusted adult or helpline if something’s worrying them.
What it won’t do
- Hand over the answer — not when pushed, not even around a worry.
- Mark, grade or rank your child.
- Guess at your child or build a hidden profile.
- Monitor or alert on them, or do anything behind their back.
- Sell their data, or use their work to train AI.
Questions parents ask
Straight answers.
Can’t my child just use ChatGPT instead?
They could — and a free “study mode” is one tap from normal mode that just gives the answer. Bearings has no off switch, and it gives you a view of their learning the free tools don’t.
Does it just refuse to help?
No — it gets them unstuck: asks what they’ve tried, points to the next move, works through the tricky part. It won’t type the answer, but it’s genuinely useful.
What about my child’s data?
No behavioural profiling, no ads, never sold, and never used to train AI. You can erase everything with one request.
Which ages and which curriculum?
Ages 8–18, built around your child’s own curriculum — American (Common Core, AP, and the Workshop model) or English (national curriculum and GCSE).
Is the parent view spying on my child?
No. Your child sees exactly what you see — nothing is hidden from them. It’s insight, not surveillance.
Why we built it
Because “they just put it in to get the answer” should worry all of us.
The most common thing children use AI for at home is homework — and the thing parents say, again and again, is that their child isn’t learning, they’re just copying. We didn’t want to ban AI from our kids’ lives; we wanted one they could actually use and we could actually trust. So we built the thing the chatbots don’t have: a tutor that can’t hand over the answer — and a way for you to see the learning is real.
No more wondering whether the AI did it for them.
Your child gets genuinely better at the work — and you get to watch it happen. That’s the whole point.
Hi — I’m your tutor.
Tell me what you’re working on and we’ll figure it out together. I won’t just give you the answer — I’ll help you find it yourself, in a way that clicks for you.
A scripted demo of how the tutor behaves — the real tutor uses these same modes. Try the options above, including “just give me the answer”.
Setting up your family
You set it up, and you stay in control.
This is where you let your child use the tutor with confidence. You set up the account and you give consent — a child can never consent for themselves. No dark patterns, no jargon.
What you’re setting up
A tutor that helps — without doing it for them.
- It can’t just hand over the answer. A separate check stops it — no off switch like the chatbots have.
- It teaches through what they love. Whatever your child chooses, never what it guesses.
- It knows their curriculum. American or English — the work their school actually expects.
- You can see they’re really learning. And your child sees it too. Insight, not surveillance.
- No scores, no profiling, no ads. Ever. You stay in control and can erase everything.
What you’ll see — rolling out
A calm, honest picture — never a report card.
This view is still being built. Here’s what it will (and won’t) show: what your child worked on, what’s worth another look, and roughly how much help they needed — never a score, grade or ranking. Your child sees exactly the same thing.
Maya · this week
A preview — illustrative only
Worked on
Worth another look
How much help
Mostly on her own this week — a few nudges on the trickier questions. (In plain words, not a score.)
Two-way by design: Maya sees this too — it’s insight you share with her, not a file on her.
A few calm steps
Confirm consent for your child.
To protect your child, we record a responsible adult and how their consent was verified before any tutoring begins. (Demo — no data is collected here.)
Find someone to talk to
If something is worrying you, you can always ask for help — and you’re never in trouble for asking.
If you’re in danger right now
Call 999 and ask for help. If you can’t speak, you can still call.
Talk to someone, free and private
- Childline — 0800 1111, free, any time, day or night.
- Samaritans — 116 123, any time, for anyone who’s struggling.
Tell a grown-up you trust
A parent, carer, teacher or counsellor can help you work out what to do next. You don’t need the right words — just say something feels wrong.
This page doesn’t record anything or ask you any questions.
How it differs
Every other tool asks one AI to behave. We put a check between your child and the answer.
Here’s how each one actually works when a child wants the answer — in plain steps. The difference isn’t a slogan; it’s where the “don’t give the answer” lives.
A normal chatbot
The catch: it just gives the answer.
ChatGPT Study Mode
The catch: coaching is a setting on the same AI — one tap from the answer, with no separate check. (As OpenAI has documented; we verify before relying on it.)
Gemini Guided Learning
The catch: a mode on one model, no separate check. (And not available to UK child accounts.)
Flint (a school tool)
The catch: teacher-set instructions on the same AI — and it’s built for schools, not parents.
Bearings
The difference: a separate check sits between your child and the answer — not a setting on the AI, and it can’t be switched off. (Plus a parent view of real learning — rolling out.)
Comparisons reflect each product as publicly documented (accessed June 2026). This is a fast-moving area — we date our claims and check them before we rely on them.
Data & child protection
Straight talk on how we handle your child’s data — and how we don’t.
We’re a trust brand, so here’s the honest version: what we collect, what we never do, where we’re careful, and what we’re still finishing before launch. No jargon, no over-claiming.
What we collect — and deliberately don’t
- Only what the tutor needs to coach your child. No home address, no school records, no behaviour profile.
- Your child’s interests are theirs to choose — we never guess them or build a hidden profile.
- Your billing details are about you, the parent, and are kept walled off from your child’s learning.
- Profiling is off by default. We don’t switch it on to target or nudge your child.
The AI behind it — the honest bit
To coach your child, their work is sent to our AI provider, which processes it outside the UK under a specific legal transfer safeguard. We won’t pretend otherwise — “your data stays in the UK” wouldn’t be true.
- The provider works only on our instructions, as a processor.
- Your child’s work is never used to train AI.
- The provider doesn’t keep your child’s work — apart from limited safety-related retention (up to about two years) needed to detect misuse. We won’t call that “zero retention”, because it isn’t.
Consent — you hold it
- Only a parent or carer can set up the account. A child can never sign themselves up or consent for themselves.
- We confirm you’re a responsible adult (a payment-card signal and a confirmed email). Honestly: no online check is perfect, and we can’t see who’s at the keyboard during a session.
- Anything optional is opt-in, never bundled with signup, and you can withdraw it as easily as you gave it.
Deleting your data — honestly bounded
You can ask us to erase your child’s data, and we’re built to honour it everywhere we hold it. We won’t claim “we delete everything instantly”, because there are a few things the law makes us keep for a while:
- Limited safety records at our AI provider (up to about two years).
- A child-protection record, if one was ever raised, under its own lawful rules.
- Tax and billing records we’re legally required to keep.
- Backups take a short, defined window to cycle out — and we re-apply the deletion if one is ever restored.
The guarantees built into the product
These aren’t lines in a policy — they’re automated checks in our code that block a release if they fail.
- No scores, grades, ranks or bands about your child — the tutor coaches, it doesn’t grade.
- It never builds or keeps a profile guessing your child’s health, ability or special needs.
- No selling, no ads, no behavioural profiling for marketing.
- Billing and learning data are firewalled — a billing record can never identify a child.
Your rights
You can ask for a copy of the data, correct it, restrict or object to its use, or have it erased — and we respond within a month. You can also complain to the ICO. A data export hides other children’s details.
Child protection & safeguarding
If your child says something worrying, the tutor calmly points them to real help — Childline (0800 1111), 999, Samaritans (116 123) — using a fixed, built-in response, not the AI. Please read this part closely, because we won’t over-promise:
- It is a signpost, not a monitoring service. It does not watch your child or scan for danger.
- It does not alert you automatically — and where a concern could involve an adult at home, it’s built so it never tips them off.
- A concern is walled off from anything you see as a parent.
- Today this is calm, always-available signposting. We’re honest that it isn’t a safety net that catches everything.
Insight, not surveillance
The parent view of your child’s learning is built on one principle: your child sees exactly what you see. It’s insight you share with them, never a hidden file on them. (This view is still being built and isn’t live yet — we’ll only switch it on once it’s been independently checked.)
Where we stand
Bearings is built to align with UK GDPR and the ICO’s Children’s Code (the Age Appropriate Design Code). We say “built to align with” on purpose — we won’t claim to be “compliant”, “certified” or “approved”, because that’s a judgement for a qualified solicitor and a regulator, not a marketing line.
We’re pre-launch. The protections above are designed and built into the product; the final legal sign-off — and a few items like the live parent view and a named data-protection officer — are being completed before we open to real families. We’d rather tell you that than pretend we’re further along than we are.
Worried about something right now? Find someone to talk to →