The sharpest question a parent can ask about aitutors.me: "My child already has Claude — why pay £14 a month?" It deserves an honest answer, and the honest answer starts with an admission: we are not smarter than Claude. We are Claude, wrapped and constrained. If you're expecting a different model, stop reading — you'll be disappointed.
What we're actually selling is the wrapper. And the question is whether the wrapper matters.
At a glance
| Raw Claude | aitutors.me | |
|---|---|---|
| Underlying model | Claude (Anthropic) | Claude (Anthropic) — same model |
| Default behaviour on "just tell me the answer" | Gives the answer | Refuses — enforced by code, not prompt |
| Cross-session learner memory | None — each chat is fresh | Per-topic mastery, pitfalls, spaced review |
| Wellbeing and safeguarding | General content safety | Energy check, Childline hard-fail, parent alerts |
| UK KS3 curriculum scaffolding | Helpful but generic | Curated hint-ladders, misconception library |
| Parent visibility | None | Session summaries, dashboard, wellbeing alerts |
| Price | Free tier / Claude.ai subscription | £14/month Founding (first 100), £24/month after |
| Anti-cheat posture | Answer-giver by default | Answer-withholder by design |
The honest starting point
We run on the parent's own Claude account. That means the model intelligence is identical whether your child opens the chat box directly or connects through aitutors.me. If anyone tells you aitutors.me has a smarter or more capable underlying model, they're wrong.
So the entire value proposition reduces to one question: what does the wrapper add that raw Claude structurally cannot do by default?
If the answer were "a nicer Socratic persona prompt," there would be no business worth having. A sufficiently determined parent — or a Year 9 student who's watched one YouTube video on prompt engineering — can replicate a persona prompt in about four minutes. The real answer is four things the raw chat cannot do at all.
What the wrapper actually adds
1. Code-enforced "never give the answer"
Raw Claude is trained to be helpful. When a student types "just tell me the answer," helpful Claude tells them the answer. It might add a worked explanation. It might express mild reluctance. But it gives them the answer.
aitutors.me's refusal does not live in a system prompt. It lives in a code layer that is regression-tested before every deploy — if a deploy breaks the "never give the final answer" rule, the deploy is blocked. No 13-year-old can talk their way out of it, because there is no conversational path to the answer. The most they can get is a fourth-level hint that makes the thinking as clear as possible without doing the thinking for them.
This is not a subtle difference. A homework assistant that folds under pressure is a homework machine. A tutor that structurally cannot fold is a tutor.
2. Persistent per-learner memory
Each Claude conversation starts fresh. There is no record of what your child got wrong last Tuesday, no tracking of which topics they consistently drop the second term in, no spaced-review queue that surfaces a half-understood concept three weeks later when it is about to matter for a test.
aitutors.me maintains a per-student database: topic mastery levels, recurring pitfalls, a structured spaced-review schedule. Professor Pi (the KS3 maths tutor) knows what your child struggled with last session and will re-test those specific gaps — not by coincidence, but by design. That is a database engine, not a prompt.
3. Wellbeing and safeguarding infrastructure
Raw Claude has responsible AI guardrails. Those are not the same as a duty of care.
aitutors.me's Mentor agent opens every session with an energy check (GREEN / AMBER / RED) and adjusts session intensity accordingly. If a student shows a sustained low-energy trend, a parent alert email is triggered. If a student shows any sign of distress or mentions something safeguarding-relevant, the system hard-fails to a single message: "If you're struggling, please talk to a parent or call Childline 0800 1111." That rule is in code, it is tested, and it cannot be suppressed by a student conversation.
None of that is in the default Claude interface.
4. Curated KS3–GCSE pedagogy
Raw Claude knows a great deal of mathematics. It does not consistently apply a 4-level hint ladder pinned to specific KS3 strand objectives. It does not have a misconception library for Year 8 algebra, curated from the specific errors UK students repeatedly make. It will not walk a student through the CPA (Concrete–Pictorial–Abstract) approach if that is what the child needs.
aitutors.me's curriculum scaffolding is content IP: expert-authored, curriculum-mapped, and consistent session to session. That consistency is part of what makes it a tutor rather than a helpful assistant.
Where raw Claude is genuinely better
It is worth being fair here.
Breadth. Raw Claude can help with every subject, every exam level, creative writing, coding, history, foreign languages. aitutors.me Phase 1 is KS3 maths. If your child needs help across six subjects, raw Claude covers more ground today.
Flexibility. A homework session that wanders into interesting tangents, a creative writing project, a revision session that mixes history and geography — raw Claude can do all of that fluidly. A disciplined Socratic tutor is the right tool for structured practice, not every use case.
Cost. If your child already has a Claude subscription, the marginal cost of using it for occasional homework help is zero. aitutors.me is £14/month on top of that.
The moat question: can a parent replicate this with a prompt?
This is the right question to ask. The honest answer: partly.
A parent can write a system prompt that asks Claude to be a Socratic tutor. That prompt will work reasonably well for an afternoon. It will not:
- Stop working when the child rephrases the question twelve times.
- Remember what happened in last week's session.
- Email the parent when the child seems distressed.
- Be regression-tested to prevent regressions in tutoring discipline.
The persona is replicable. The enforcement, the memory, the parent loop, and the curriculum content are not — they require software infrastructure that is meaningfully more than a prompt.
Who should just use raw Claude
- Your child uses AI across many subjects and you are comfortable with answer-giving as a study mode.
- They are self-directed enough to choose "explain this, don't just tell me."
- Maths is not the subject where they struggle most right now.
- Budget is a real constraint.
Who should use aitutors.me
- You are specifically worried about AI doing your child's homework for them.
- You want a parent alert if something seems wrong with your child's wellbeing.
- Your child is working through KS3 maths and you want consistent, disciplined practice.
- You want visibility into how sessions are going without sitting next to them.
FAQ
Does aitutors.me use Claude?
Yes. aitutors.me is built on Anthropic's Claude — the same model your child would access via claude.ai. The difference is entirely in the wrapper: disciplined Socratic enforcement, persistent learner memory, and a parent-alert loop.
Can't I just tell Claude to be a Socratic tutor?
You can write that prompt. The problem is that a 13-year-old's whole incentive is to argue past it — and Claude, being helpful by default, usually folds. aitutors.me's refusal is in code, not in a prompt the child can debate.
Is aitutors.me more expensive?
Yes, by £14/month (Founding price). That buys the disciplined wrapper, not a different model.
Why does the wrapper matter?
Because a tutoring tool that gives the answer when pressured is not a tutoring tool — it is a homework shortcut. The wrapper is the product.
Is aitutors.me's data safe?
Conversations are stored on EU infrastructure under UK GDPR, with a 30-day rolling deletion policy. No third-party model training on conversation data.
Written by Jason (founder of aitutors.me). I have a financial interest in aitutors.me. I do not have any financial relationship with Anthropic. Updated 1 June 2026.