Ask a Year 9 class where they would go to check a claim. A teacher? A search result? An AI chatbot? In a new survey of 4,021 young people aged 11 to 24 in England, more chose AI than a person to check a fact: 47% against 21%.
That is the number I would put in front of every subject department. A confident explanation and a confident verification can come from the same system, and a pupil may experience that as two pieces of evidence.
The survey, conducted by Public First in August for a September report by Sir Anthony Seldon and Tim Bunting, found that 87% of young people had used an AI tool. Only 9% thought schools and universities were preparing them very well for a world with AI. Two-thirds said teaching responsible use was the job of those institutions.
The gap is not about whether pupils can find the button. They already have.
The answer that checks itself
Imagine a student asks a chatbot for a paragraph on the causes of the First World War. It gives them a neat account. They ask the same chatbot whether the paragraph is accurate. It says yes, perhaps with a few minor corrections. The student has apparently checked their work without leaving the original source of the claim.
History teachers know what is missing. Where did each claim come from? Which interpretation is contested? What would an independent source say? The AI literacy here is ordinary historical method applied to a new kind of text. The same pattern appears in science when a model explains an experiment and then judges whether its explanation was sound.
The survey does not tell us how often young people actually verify claims this way. It reports whom they say they trust. That is enough to give schools a teaching question: do pupils understand the difference between asking a system to sound certain and checking a claim against evidence?
Access is uneven too
Public First found that 92% of young people in the highest social grade had used AI, compared with 79% in the lowest. One in five young people who were not in education, employment or training had never used an AI tool. These figures describe access and use, not skill. They do, however, challenge the assumption that leaving AI learning to home experimentation gives everyone the same start.
The report makes much larger claims about redesigning education and sets out a ten-point policy plan. Some of those proposals will divide opinion. The survey result is useful on its own. Schools can teach pupils how to corroborate a generated answer while that national debate continues.
A lesson you can run this term
Choose a question your class already has enough subject knowledge to investigate. Ask an AI tool for a short answer. Give pupils two independent sources and ask them to annotate every claim in the answer: supported, disputed, or still unverified. Then ask the tool to check its own answer and compare its verdict with theirs.
The aim is to make visible what checking actually involves: independent evidence, subject knowledge, and the willingness to leave a claim unresolved. The pupil who can explain why the chatbot's second answer is not a second source has learned something that travels well beyond one product.
Nearly nine in ten young people in this survey have tried AI. The urgent question for schools is whether any of them have been taught to distrust an answer at the right moment.
Sources: Public First, Young People's Views on AI and Education, 22 September 2026; Seldon and Bunting, Britain's Greatest Education Opportunity Ever. Survey: 4,021 people aged 11–24 in England, 18–27 August 2026.