The Tutor Nobody Mentions at the School Gate
Almost half of London's secondary pupils have had a private tutor, and hardly anyone talks about it. The strange part is that tutoring really is one of the best-evidenced things in education - just not in the form most families are buying.
Ask around at pickup and almost nobody admits to having a tutor. The figures say something else. Nearly three in ten secondary pupils in England and Wales have had private tutoring at some point, according to the Sutton Trust's report published this February. Twenty years ago it was fewer than one in five.
In London it is 45 per cent. In rural areas, 19 per cent. Same curriculum, same exam papers, wildly different amounts of teaching going on after the bell.
What makes this awkward is that tutoring is not snake oil. It is one of the most solidly evidenced things in education. The awkward bit is that the version with the evidence behind it is not really the version most families are buying.
Then you read the small print. Sessions of about half an hour, three to five times a week, running for something like ten weeks. Tied directly to what the class is covering that term. Aimed at a child who has fallen behind on something specific, with somebody checking whether it is working.
That is a different animal from an open-ended hour on a Sunday afternoon that carries on for two years because stopping feels risky. When researchers look at private tutoring as families actually buy it, the results get much messier. Some studies find decent gains. Others find almost nothing. What separates them is usually whether the child was genuinely behind, and whether the tutor had any idea what they were doing.
The NFER evaluations were sobering. School-led tutoring produced roughly one month of extra progress in maths, both at the end of primary and at GCSE. In English it was less than a month. Real, but a long way short of five.
What mattered turned out to be dose and familiarity. Benefit seemed to need more than about twenty hours per pupil, and the strongest version was the one run by a child's own school, often by adults who already knew them. Six sessions with a stranger on a screen moves very little.
That funding has now ended. Last year 58 per cent of schools cut back what tutoring they offered, and only one in five pupils gets any through school, slightly down on 2023. The public version is shrinking at exactly the moment the private one is growing.
The sharper divides are geography and community. A third of urban pupils have been tutored against a fifth of rural ones. By ethnicity the differences are enormous: 64 per cent of Black pupils, 50 per cent of Asian pupils, 20 per cent of White pupils. In the poorest areas the contrast widens rather than closing, with 65 per cent of Black pupils tutored and 10 per cent of White pupils.
So this is partly about money and partly about something else. Whether the families around you treat an extra hour of maths as a normal expense, or as something other people do. Around four in ten teachers have worked as private tutors themselves, which tells you how much of this runs on word of mouth.
Most tutoring is crammed into Year 11, in the months before GCSEs. Going by the research on timing, that is close to the least useful moment to start.
A Harvard trial published in Scientific Reports last June gave 194 physics undergraduates either a purpose-built AI tutor or their usual class. The class was already an active-learning one, the sort that beats lectures. The AI tutor group learnt roughly twice as much, in less time. A UK maths trial found an AI tutor working alongside human supervision matched what expert human tutors achieved. A GCSE science trial this year beat ordinary revision fairly comfortably.
The caution comes from the failures. When researchers examined Khan Academy's AI tutor, they found it did not adapt to the individual child and did nothing to help them think about their own learning. The Harvard version had been hand-tuned to keep its answers short so students were not swamped. The words "AI tutor" tell you nothing at all. The design tells you everything.
The evidence says short and targeted beats long and vague, and that the best person for the job often already works at the school. Worth asking before you go looking on a booking site.
There is a stranger thing going on underneath all this, though. Once enough families are running a second timetable, it stops counting as extra. It quietly becomes the level. And the children without one are not falling behind exactly. They are being measured against a class that simply had more lessons than they did.
In London it is 45 per cent. In rural areas, 19 per cent. Same curriculum, same exam papers, wildly different amounts of teaching going on after the bell.
What makes this awkward is that tutoring is not snake oil. It is one of the most solidly evidenced things in education. The awkward bit is that the version with the evidence behind it is not really the version most families are buying.
What the evidence is actually about
The Education Endowment Foundation puts one to one tuition at around five extra months of progress. That is a large claim for something so simple, and it holds up across a lot of studies.Then you read the small print. Sessions of about half an hour, three to five times a week, running for something like ten weeks. Tied directly to what the class is covering that term. Aimed at a child who has fallen behind on something specific, with somebody checking whether it is working.
That is a different animal from an open-ended hour on a Sunday afternoon that carries on for two years because stopping feels risky. When researchers look at private tutoring as families actually buy it, the results get much messier. Some studies find decent gains. Others find almost nothing. What separates them is usually whether the child was genuinely behind, and whether the tutor had any idea what they were doing.
When the country tried to buy it for everyone
After the pandemic, England tried to hand tutoring to every child who needed it. The National Tutoring Programme cost well over a billion pounds and was, on paper, exactly what the evidence recommended.The NFER evaluations were sobering. School-led tutoring produced roughly one month of extra progress in maths, both at the end of primary and at GCSE. In English it was less than a month. Real, but a long way short of five.
What mattered turned out to be dose and familiarity. Benefit seemed to need more than about twenty hours per pupil, and the strongest version was the one run by a child's own school, often by adults who already knew them. Six sessions with a stranger on a screen moves very little.
That funding has now ended. Last year 58 per cent of schools cut back what tutoring they offered, and only one in five pupils gets any through school, slightly down on 2023. The public version is shrinking at exactly the moment the private one is growing.
Who is actually buying it
Not simply the wealthy, which is the first surprise. Thirty per cent of pupils from the best-off homes have had a tutor, against 23 per cent from the worst-off. A real gap, but narrower than you would expect for something costing about £37 an hour, and considerably more in London.The sharper divides are geography and community. A third of urban pupils have been tutored against a fifth of rural ones. By ethnicity the differences are enormous: 64 per cent of Black pupils, 50 per cent of Asian pupils, 20 per cent of White pupils. In the poorest areas the contrast widens rather than closing, with 65 per cent of Black pupils tutored and 10 per cent of White pupils.
So this is partly about money and partly about something else. Whether the families around you treat an extra hour of maths as a normal expense, or as something other people do. Around four in ten teachers have worked as private tutors themselves, which tells you how much of this runs on word of mouth.
Most tutoring is crammed into Year 11, in the months before GCSEs. Going by the research on timing, that is close to the least useful moment to start.
The tutor that will soon be free
In January the government announced plans to put AI tutoring tools into schools, aimed at disadvantaged pupils, with a pilot due in 2027. The obvious reaction is scepticism, and the evidence is more encouraging than the scepticism allows.A Harvard trial published in Scientific Reports last June gave 194 physics undergraduates either a purpose-built AI tutor or their usual class. The class was already an active-learning one, the sort that beats lectures. The AI tutor group learnt roughly twice as much, in less time. A UK maths trial found an AI tutor working alongside human supervision matched what expert human tutors achieved. A GCSE science trial this year beat ordinary revision fairly comfortably.
The caution comes from the failures. When researchers examined Khan Academy's AI tutor, they found it did not adapt to the individual child and did nothing to help them think about their own learning. The Harvard version had been hand-tuned to keep its answers short so students were not swamped. The words "AI tutor" tell you nothing at all. The design tells you everything.
The question worth asking first
If you are weighing it up, the useful question is narrower than "should we get a tutor". It is: what, precisely, is my child stuck on? Fractions with different denominators. Structuring a paragraph. Reading a graph. If you cannot name it in a sentence, the tutor will not be able to either, and what you are buying is mostly reassurance for yourself.The evidence says short and targeted beats long and vague, and that the best person for the job often already works at the school. Worth asking before you go looking on a booking site.
There is a stranger thing going on underneath all this, though. Once enough families are running a second timetable, it stops counting as extra. It quietly becomes the level. And the children without one are not falling behind exactly. They are being measured against a class that simply had more lessons than they did.
Sources & Further Reading