Ask a fourteen-year-old to check whether something is true, and there is now a better-than-even chance they will not ask a person.

New polling of 4,021 people aged 11 to 24 in England, conducted between 18 and 27 August 2026 by Public First, found that when asked who they trust to check a fact, 47 per cent said AI and 21 per cent said a person. Among young men it rises to 52 per cent.

That number travelled, and deservedly. It is also not the most important thing in the dataset.

87 per cent have used an AI tool. Usage peaks at 90 per cent among 13 and 14-year-olds - younger than most parents assume, and younger than most school policies are written for.

So this is not an emerging trend to get ahead of. It is the current baseline.

38 per cent agree they are better at using AI than their teachers.

Only 9 per cent think schools and universities are preparing them very well for a world with AI - while two-thirds believe it is the school's job to teach them to use these tools responsibly.

Read those as a pair and you get something more specific than "kids think school is rubbish." They are not rejecting the institution's authority. They are still assigning it the responsibility. They simply do not think it is being discharged well, and a substantial minority think they have already overtaken the adult in the room on the relevant skill.

That is a school that has lost informational authority before it has written its rules - not because anyone argued it out of the position, but because the default tool for checking things changed and the institution did not move at the same speed.

You will see this one everywhere: "42 per cent say AI has made school less important."

True, and half a result.

The actual finding is 42 per cent said less important, against 35 per cent who said more important.

A seven-point gap is a split, not a verdict. Young people are genuinely divided on whether AI raises or lowers the value of school, and a headline built on the 42 alone is telling you about one side of a near-even argument. We are flagging it because it is the most quotable and most distorting figure in the release.

One more piece of care: that 9 per cent is a top box. The question asked about being prepared very well. We do not have the "fairly well" number from the public summary. So 9 per cent is not "91 per cent think school is failing them." Keep the word "very" attached to it wherever it goes.

Well down the list, without a headline:

Two in five. For problems and feelings.

This is the same behaviour a JAMA Pediatrics study found among North American adolescents earlier this year - a substantial minority turning to chatbots for emotional support rather than homework. Two independent datasets, two countries, one pattern.

The public conversation about AI and young people is still largely a conversation about cheating on essays. Meanwhile something considerably more consequential is already at scale, and it is happening with a tool that has no duty of care, no training, no ability to notice a pattern across weeks, and no one to escalate to.

We are not saying it is all harmful. A young person rehearsing how to describe a worry before telling a human being is doing something useful. The problem is that nobody designed it for this, and nobody is checking.

One result complicates the story, and it should be in every honest summary of this poll.

92 per cent of young people from the highest social grade have used AI. 79 per cent from the lowest. And 1 in 5 NEET young people - not in education, employment or training - have never used an AI tool at all.

The dominant anxiety is that children are using too much AI. For the most disconnected group in this survey, the live risk may be the opposite one: that they arrive in a labour market reshaped by a tool they have never touched, while their better-off peers have had three years of practice.

Both problems are real. They need opposite responses, and a policy built only on the first one will widen the second.

Also worth noting: 71 per cent expect there will be fewer jobs because of AI. That is the backdrop these young people are answering every other question against.

The numbers above are Public First polling - a member of the British Polling Council, nationally representative, decent sample. We have no quarrel with the data.

The report it appears in is a different kind of object. "Britain's Greatest Education Opportunity Ever" is an advocacy document that ends in a Ten-Point Plan for Government. Its authors are Sir Anthony Seldon - former headmaster of Brighton, Wellington and Epsom Colleges, founding Director of Wellington College Education, and the report includes a section on AI in schools written from the Wellington perspective, his own institution - and Tim Bunting, co-founder of the venture capital firm Balderton Capital and still a senior adviser there.

None of that makes the polling wrong. It does mean the argument around the polling is being made by people with positions, and you are entitled to know what they are before you accept the conclusions. Take the numbers; evaluate the plan separately.

One last limit, which applies to the whole thing: this is attitudes, not outcomes. Every figure is what young people say about themselves at one moment in time. None of it measures whether AI has actually helped or harmed anyone's learning.

Make verification a visible habit, not a lecture. The 47/21 split is not evidence that young people are gullible. It is evidence that they have picked the tool that is fastest, always available, and never makes them feel stupid for asking. You will not beat that with a warning. You beat it by checking things out loud in front of them, including things the AI told you.

Ask what it's being used for, not how much. Screen time is the wrong metric here. "Have you ever asked it about how you were feeling?" is a different and much more informative question - and, given two in five, a likelier yes than you expect.

Don't make the emotional use shameful. If a child is using a chatbot to work out what is bothering them, the outcome you least want is that they hide it. The aim is that it becomes a step toward a person, not a substitute for one.

Ask the school what its actual rule is. Two-thirds of young people say it is the school's job. Nine per cent say it is going very well. There is a reasonable chance your child's school has no written position at all, and it is a fair question at the next parents' evening.

Watch for the opposite problem too. If your child is the one not using these tools, that is no longer automatically the safe side of the line.

They are not confused about what a chatbot is. They have made a choice about who to ask, and by more than two to one they are not choosing us.

The useful response is not to win the trust back by decree. It is to be the person who shows them how to check.

Source: Sir Anthony Seldon and Tim Bunting, "Britain's Greatest Education Opportunity Ever," Public First, September 2026. https://www.publicfirst.co.uk/seldon-bunting-report

Polling: Public First, a member of the British Polling Council. Nationally representative survey of 4,021 people aged 11-24 in England, fieldwork 18-27 August 2026. All figures are self-reported attitudes; no causal claims are available from this design. The report containing the polling is an advocacy document with a Ten-Point Plan for Government; its authors' positions are disclosed above. The comparison to the JAMA Pediatrics findings, the reading of institutional authority, and the guidance for parents are ours.

Disclosure: drafted with Claude, made by Anthropic.

If you or your child is in crisis: in the US and Canada, call or text 988. In the UK, call 111.

Want practical AI guidance for parents and educators every week? Subscribe: https://www.aibyage.com/?modal=signup&utm_source=beehiiv&utm_medium=newsletter