You Do Not Get Fit by Watching Sport: Why AI That Makes Revision Easier Can Make GCSE Results Worse
TL;DR: Sir Anthony Seldon, former headteacher and co-founder of AI in Education, has warned that the educational promise of AI is being squandered and that the technology poses the biggest challenge yet to young people's learning and wellbeing. The most useful idea in the coverage is a simple one: you do not get fit by watching sport, and you do not learn by consuming content. Learning happens through effort, and AI tools that remove the effort raise the quality of the homework while lowering the quality of the learning. For a GCSE student that is a trap, because the exam is the one place the tool cannot come. This article explains what "cognitive struggle" means in practice, how to tell whether your child is using AI to learn or to avoid learning, and how to keep the effort where it belongs.
Key Takeaways
Learning is built by effort. Any tool that removes the effort from a task removes the learning with it.
The test is simple: could your child do the work again tomorrow without the tool? If not, nothing was learnt.
Use AI after an attempt, never instead of one. It can explain a mistake. It should not prevent the mistake being made.
Retrieval, timed exam questions and marking against a mark scheme remain the methods that move grades, because they are effortful.
Watch for homework that suddenly looks perfect. Perfect homework and flat test scores are the signature of AI doing the thinking.
Most parents have had the same quiet moment in the last year. A piece of Maths or Science homework comes back looking better than anything the child has produced before. Neat working, correct answers, the right terminology. And the next test score has not moved at all.
That gap, between what the homework looks like and what the student can actually do, is the story behind this week's warnings about AI in education.
Sir Anthony Seldon, who ran two of the country's best-known schools before co-founding the advisory group AI in Education, has written that the technology's potential is being squandered, that the risk to intellectual rigour and to the integrity of assessment is serious, and that school leaders, teachers and parents are asking for guidance that has not arrived. Britain, he argues, sits near the bottom of the international table on AI in schools.
Whether or not the government responds, the practical question for a family with a Year 10 or Year 11 student is narrower and more urgent. How does a student use these tools without hollowing out the learning that the exam will test?
The answer starts with understanding what learning actually is.
You Do Not Get Fit by Watching Sport
One expert quoted in the same coverage, discussing the OECD's findings on how teenagers use AI, put the principle in a way worth pinning to the fridge. Watching sport does not make anyone fit. Consuming content does not make anyone learn.
Learning is the effort. It is not the thing that happens after the effort.
When a student wrestles with a rearranging question, tries a method, gets it wrong, works out why and tries again, the struggle itself is what builds the understanding. Cognitive scientists have known this for a long time. Memory forms around effort, and the harder a piece of information was to retrieve, the better it sticks the next time.
AI cuts across this in a particular way. Where it increases the struggle, by challenging an answer, asking a follow-up question, or explaining a mistake the student has already made, it helps. Where it reduces the struggle, by producing the answer, the working or the essay, the task gets done and nothing is learnt.
The unsettling part is that the two look identical from the outside. The homework is complete either way. The difference only becomes visible in the exam hall, where the student is alone with a question and there is no tool to ask.
Why This Matters More for GCSE Maths and Science
Every subject is affected, but Maths and Science are affected in a specific way, because they are cumulative and because their exams are unforgiving of gaps.
In Maths, the danger is method. A student who asks a tool to solve a simultaneous equations question sees a clean, correct method. They nod along. They feel they understand it. Then in a test, with a slightly different pair of equations, the method is not there, because it was never built. It was watched.
In Science, the danger is explanation. Six-mark questions reward a structured chain of reasoning in the right scientific language. A tool will produce a model answer instantly. A student who reads model answers learns to recognise good explanations. They do not learn to produce them, and recognition is not what the examiner is marking.
In both, the danger compounds. GCSE content builds in layers. If ratio was outsourced in Year 9, the algebra that rests on it wobbles in Year 10, and the Physics rearranging that rests on the algebra collapses in Year 11. By then the student appears to be struggling with five topics. They are struggling with one, and it was never learnt.
How to Tell Whether Your Child Is Learning or Outsourcing
You do not need to police the laptop. You need to look for a few signals, most of which are visible at the kitchen table.
Homework quality has jumped but test scores have not. This is the clearest signature. Effort produces uneven work and improving scores. Outsourcing produces polished work and flat scores.
They cannot redo it the next day. Ask them to explain yesterday's homework question, or do a similar one, without the screen. If they cannot, the task was performed, not learnt.
The working is too tidy. Real working has crossings-out, a false start, a method abandoned halfway. Working that arrives fully formed on the first line may not have come from the student.
They describe topics as "easy" but avoid tests on them. Reading a correct answer feels like understanding. The avoidance is the honest signal.
Revision has become reading. Watching explanation videos and reading AI summaries feels productive and is nearly useless on its own. Revision that works involves a pen, a blank page and a question the student has not seen before.
None of these means a child is being dishonest. Most students who use AI this way believe they are learning. That is exactly what makes it dangerous.
Not sure whether your child's revision is building anything?
Our free assessment tests what a student can actually do under exam-style conditions, without notes or tools, and shows exactly where the understanding is real and where it has been borrowed.
The Rule That Keeps AI on the Right Side
There is one rule that, if a family holds to it, captures most of the benefit and avoids most of the harm.
AI comes after the attempt, never instead of it.
In practice, that looks like this. The student attempts the question in full, on paper, before opening anything. If it is wrong, they mark it against the mark scheme first and try to find the error themselves. Only then do they ask a tool to explain the specific step they cannot see. And having read the explanation, they close it and do a fresh question of the same type from memory.
Used this way, the tool is doing the job a good teacher does at the moment of confusion: explaining the one thing the student could not work out. Used the other way, it is doing the job the student was supposed to do, and the student is watching.
A worked example. A Year 11 student is stuck on a reverse percentage question: a coat costs £48 after a 20% reduction, and the question asks for the original price. The outsourcing route is to type the question in and copy the answer, £60, with the working. The learning route is to attempt it, write £38.40, check the mark scheme, see that the answer is wrong, and only then ask why. The explanation, that £48 is 80% of the original, lands because it answers a question the student was already asking. Next lesson, a similar question on a house price comes up, and the method is there and secure.
Same tool. Opposite outcome. The difference was where the effort sat.
What Structured Tuition Does That a Tool Cannot
This is also the reason effective tuition has not been made redundant by AI, and if anything has become more valuable.
Good tuition is built around effort rather than answers. In our 3-Step framework, the first step, precision diagnosis, is deliberately done without tools: a student works through exam-style questions under timed conditions so that what they can genuinely do is separated from what they can recognise. The second step, smart revision and exam strategy, teaches the effortful methods that build retention, retrieval practice, self-marking against the mark scheme, spaced repetition, and how to write a six-mark answer from a blank page. The third step, confidence and accountability, is the human part: somebody who notices when the homework looks too good, asks the student to explain it, and keeps the effort honest week after week.
A tutor's job is not to make the work easier. It is to make sure the difficulty is the right kind, at the right level, with feedback quick enough that the struggle stays productive rather than demoralising. That is a judgement, and it is the part of teaching Seldon is right to say must not be dehumanised.
What to Do This Term
Three practical moves, none of which requires banning anything.
Agree the rule out loud. Attempt first, tool second, fresh question third. Say it once, together, and refer back to it. Students respond far better to a clear principle than to surveillance.
Add a weekly blank-page check. Once a week, ten minutes, a handful of mixed questions on recent topics, done with nothing open. The score matters less than the pattern. If topics that looked secure in homework fall apart on the page, that is the diagnosis, and it is far better to have it in October than in May.
Make revision active by default. Swap "read the notes" and "watch the video" for "cover the page and write down what you remember", "do three exam questions and mark them", and "explain this topic to me in two minutes". Each of those is effortful, which is precisely why each of them works.
The tool that does the thinking is not helping. The tool that makes the thinking sharper is.
Seldon's warning is aimed at government and schools, and it may or may not be heard there. For a family, the lesson is smaller and entirely within reach. Learning has always been effort, AI has not changed that, and the students who do well next summer will be the ones whose effort stayed their own.
That is the principle behind everything we do. 95% of our students improve by at least one grade within their first term, and our students have achieved top GCSE grades every year since 2019. You can see the full results record here.
Want an honest picture of what your child can do without the screen?
Book a free assessment and we will show you where the understanding is solid, where it has been borrowed, and how to build a revision routine that holds up in the exam hall.
About the Author
A-Star Tuitions Team
The team at A-Star Tuitions specialises in GCSE Maths and Science tuition across the UK. We share practical, evergreen insights to help our readers succeed.
