Two kinds of app
An answer machine starts from the question. Your child types it in or takes a photo, and the app returns the solution, often laid out step by step. The app has done the maths; your child reads it.
A tutor starts from your child’s own working. It looks at what they have tried, points at where it went wrong, and leaves the next step to them.
Step-by-step solutions can look like teaching, and reading one is better than copying a bare answer. But reading a worked solution is a different skill from producing one, and the exam asks your child to produce one.
Why it matters for GCSE maths
It is all exams. Whichever board your child sits (AQA, Pearson Edexcel, OCR or Eduqas), GCSE maths is assessed only by written papers, and one of them doesn’t allow a calculator. There is no app in the exam hall, so the only thing homework can carry into the exam is what your child learned doing it.
Marks follow the working. AQA’s marking guidance says method marks “are awarded for a correct method which could lead to a correct answer.” Examiners keep giving method marks after a slip, “providing the correct calculation is shown.” A child who can set out a method earns marks even when the final answer is wrong. Copying a worked solution doesn’t build that habit.
The research points the same way. A study published in PNAS in 2025 gave nearly 1,000 high school students in Turkey one of two AI tools to practise maths with. One worked like a standard chatbot. The other was designed to give hints rather than answers. Both helped during practice. In the exam afterwards, without AI, students who had practised with the standard chatbot scored 17% lower than students who had practised with no AI at all. The hints group scored about the same as the no-AI group. The authors concluded that students had used the standard chatbot as a “crutch”.
Two things are worth keeping in mind. It is one study, in one country, measured over a short period, and the authors call for longer follow-up. And the hints tool didn’t raise exam scores above practice without AI; it avoided the harm. That is a fair bar to hold any AI tool to, ours included.
AI can be confidently wrong. The Department for Education’s guidance on AI in education warns that generative AI “can produce nonsensical, inaccurate or false information presented as fact.” A student copying a solution has no way to tell when that has happened. The same guidance suggests schools may wish to review their homework policies with AI in mind, so it is worth asking what your child’s school expects.
A five-minute test for any app
Sit with your child and try one real homework question.
- Type the question in, with no working. Does the app show the whole solution, or does it ask what your child has tried?
- Give it working with a mistake on the second line. Does it find that line? Or does it ignore the working and show its own method from the start?
- Ask it outright for the answer. Then ask again. Some tools that start by giving hints will hand the answer over if asked twice.
- Ask how it knows a line is right. Is the same AI that writes the replies also deciding what’s correct, or is there a separate check?
- Ask what you get to see. Scores, time spent, the conversations themselves? Agree it with your child before they start, so nobody is surprised later.
There is no perfect score on this test. It shows you what the app does when your child is stuck and tired at 9pm, which is when it matters.
Signs an app might be doing the homework
None of these prove anything on its own, but each is worth a gentle question:
- It’s done in minutes. A 30-minute set finished in five is worth asking about.
- There’s no working. GCSE maths gives marks for method. An answer with nothing behind it is worth a look.
- They can’t explain a step. Pick one line and ask why it’s there. If they wrote it, they can usually tell you.
- Homework and tests don’t match. Full marks at home and much lower marks in class tests is the clearest sign of all.
If you see these, start with a conversation rather than a ban. Ask your child to talk you through one question they found hard, and what they do when they get stuck.
Where Mathable fits
The GCSE maths tutor that won’t do your child’s homework for them
Here is how it answers the five checks above:
- It starts from your child’s working. They type it line by line, or take a photo of the page.
- It points at the first line to fix. Then it gives one hint at a time, and your child works out the fix.
- It holds the answer back while it hints. It can show a full worked example once two hints haven’t helped, or when your child asks for one.
- Ticks come from a maths checker. Code checks every line it can read. A green tick means the checker confirmed that line is right; it isn’t the AI’s opinion.
- You see results, not conversations. You can set tests, and Mathable marks them. The Parent view shows mastery in each GCSE maths area, how often your child practised this week, and how each test went. It doesn’t show their conversations.
A free week, then £19.99 a month or £119.99 a year.
The free week needs a card. You won’t be charged if you cancel before the week ends.
Sources
- Bastani, H., Bastani, O., Sungu, A., Ge, H., Kabakcı, Ö. and Mariman, R. (2025), “Generative AI without guardrails can harm learning: Evidence from high school mathematics” PNAS 122(26), e2422633122.
- Knowledge at Wharton, “Without guardrails, generative AI can harm education” (a summary of the study above, including where it took place).
- AQA, GCSE Mathematics (8300): Specification at a glance
- Pearson Edexcel, GCSE (9–1) Mathematics (1MA1) specification Issue 2.
- OCR, GCSE (9–1) Mathematics (J560) specification Version 2.1.
- Eduqas, GCSE (9–1) Mathematics specification Version 3.
- AQA, GCSE Mathematics (8300) marking guidance
- Department for Education, “Generative artificial intelligence (AI) in education” (updated 12 August 2025).