Can You Trust What AI Tells You? What Students and Teachers Need to Know
AI can explain a difficult concept, help a student explore an unfamiliar subject or suggest ideas for an assignment. But what happens when the answer sounds convincing and turns out to be wrong? As more young people use AI for learning and information, helping them question its answers is becoming an important part of digital literacy.
How are students using AI?
The Higher Education Policy Institute’s (HEPI) Student Generative AI Survey 2026, which heard from 1,054 full-time UK undergraduates, found that 95% now use AI in at least one way.
Among children, use is lower. Ofcom’s latest research found that 56% of 8- to 17-year-olds have used AI, rising to 66% among 16- to 17-year-olds.
However, using AI does not necessarily mean relying on it for every assignment. Research by Oxford University Press (OUP) with almost 4,000 UK 13- to 18-year-olds found that one in four (24%) regularly seek out AI tools for schoolwork, while one in three (34%) only use them when a teacher suggests it.
These findings show that AI is becoming part of young people's lives, but how they use it varies. Understanding when AI can help and when its answers need checking is increasingly important.
Why can AI sound right when it is wrong?
To understand why AI can get things wrong, it helps to know how chatbots produce their answers.
Tools such as ChatGPT, Copilot and Gemini use large language models, which are trained on vast amounts of data, including text. When you ask a question, the model generates a response by predicting what is likely to come next, based on patterns it has learned. Some AI tools can also search the web or draw on other sources to help answer questions.
A fluent, confident response is not necessarily an accurate one. AI can produce misleading information, invent details or misrepresent its sources. This is why its answers need to be checked rather than accepted at face value.
How often does this happen? A large study coordinated by the European Broadcasting Union (EBU) and led by the BBC investigated the problem. Journalists from 22 public service media organisations in 18 countries reviewed more than 3,000 responses from ChatGPT, Copilot, Gemini and Perplexity to questions about news and current affairs.
Almost half (45%) had at least one significant issue. A third had serious sourcing problems, such as missing, misleading or incorrect attribution, while one in five contained major accuracy errors, including invented details and out-of-date information.
Published in October 2025, the study focused on news and current affairs rather than homework. Although the results showed some improvement compared with the BBC's earlier research, significant problems with accuracy and sourcing remained.
This matters for schools because young people are also turning to chatbots for news. The Reuters Institute's Digital News Report 2025, based on a survey of almost 100,000 people in 48 countries, found that 15% of under-25s use chatbots for news every week.
These findings do not tell us how often AI gets homework questions wrong. They do, however, highlight the need to check AI-generated information, particularly when it concerns current affairs or unfamiliar subjects.
But if AI can produce convincing yet inaccurate answers, can the people using it recognise when something is wrong?
Can students tell when AI has got something wrong?
Research into whether young people can reliably identify errors in AI-generated information is still developing, particularly for secondary-school pupils in the UK.
Ofcom's research with children found that many did not appear to distinguish between different AI tools, and few questioned their accuracy. Four in ten teenage AI users aged 13 to 17 said they would trust an AI-generated news article as much as, or more than, one written by a human, while 43% said they would trust it less.
These figures measure trust in news, not the ability to spot factual errors, so they only tell us part of the story.
Studies that directly test whether students can detect AI errors are mostly small and focused on university-level learners. A UK business school study involving 211 second-year students found that the ability to detect AI errors was linked to academic ability and critical thinking. However, the study is a preprint and has not yet been peer-reviewed.
More encouragingly, a peer-reviewed study of middle-school learners who built their own chatbots found that young people can develop strategies to identify AI errors when given appropriate support.
What is missing is a large UK study of secondary-school pupils that establishes how well they can identify inaccurate AI-generated information. Until more evidence is available, teachers can start by asking their own classes: do students check what AI tells them, and how can they develop this habit?
Recognising mistakes is one challenge. Another is knowing when it is acceptable to use AI for schoolwork.
Is using AI for homework cheating?
Students themselves are unsure. In OUP's research, 44% thought it was cheating to use AI to complete all their homework, and almost one in five thought it was cheating even to ask an AI tool for homework tips. Just 15% said they had been given enough guidance on AI.
The Joint Council for Qualifications (JCQ), which represents the exam boards, says that work submitted for qualifications must be the student’s own, that any use of AI must be acknowledged, and that students who misuse AI to the extent that their work is not their own will have committed malpractice.
These requirements concern formal assessments, including coursework and other non-examined assessments. Policies on everyday homework can vary between schools, which may help explain why students feel uncertain about what is acceptable.
What students say they want from AI is revealing. Asked what helps them learn most effectively, 44% preferred tools that suggest tasks to help them understand a subject, while 41% preferred tools that ask questions to help them reach an answer independently. Only around one in five wanted the answer straight away.
For schools, the question is not simply whether using AI is cheating. It is also about what students should be able to understand and do for themselves.
What are the arguments for and against AI in schools?
Supporters point to the time AI can save and its potential to make difficult ideas easier to understand. Among the teenagers in OUP's research, 39% were more excited than worried about AI's impact on their education, compared with 16% who were more concerned. In addition, 77% would like their teachers to use AI in lessons, for example to make complex work easier to understand.
However, there are real concerns. HEPI's report highlights worries that AI could undermine independent thinking, fairness and academic skills. The accuracy problems discussed above also raise questions about how confidently students should rely on AI-generated information.
The Department for Education (DfE) acknowledges that there is limited evidence about how AI use affects learners' development and educational outcomes.
The debate is less about whether AI is good or bad than about when it supports learning and when it replaces it. AI may help students explore unfamiliar topics or approach difficult tasks differently. But if they accept its answers without checking them, they risk missing errors and losing opportunities to develop their own understanding.
AI literacy: the common ground
Views on AI in schools differ, but one point keeps emerging: young people need to learn how to question the information AI provides.
This is part of a wider conversation about media literacy and online safety. In our post on the under-16 social media ban, we discussed the Guardian Foundation's argument that any ban must be accompanied by properly funded media literacy education.
The same principle applies to AI. Setting rules for its use is only part of the picture. Students also need the skills to assess information, recognise when it may be misleading and find reliable sources.
HEPI's report argues that universities should build AI literacy into their curricula. It also found that students overwhelmingly see AI skills as essential for their futures, yet fewer than half feel supported by their teaching staff.
Developing these skills earlier can help young people approach AI with greater confidence and critical awareness.
Why this matters for schools
Whatever schools decide about AI policy, young people need the skills to question and evaluate the information they encounter online.
PSHE and Citizenship lessons offer a natural opportunity to develop these skills. One practical activity is to ask an AI tool a question the class already knows the answer to, then examine its response together.
Are the facts correct? Can its claims be verified using reliable sources? Does it acknowledge uncertainty, or present everything with the same level of confidence?
Students can then apply the same questions to a topic they are less familiar with. This helps them practise checking information rather than accepting a convincing answer at face value.
Schools should also ensure that classroom activities follow their own policies on AI use, data protection and safeguarding. The DfE provides guidance and support materials on generative AI in education to help schools use these tools safely and effectively.
Explore the Issues series: resources for teaching AI and digital literacy
Helping students question AI-generated information is part of a wider conversation about online safety, digital literacy and responsible technology use. The following titles from Independence Educational Publishers support classroom discussions around these themes.
Vol. 409: Online Safety explores both the benefits and risks of life online, helping students develop the awareness they need to make informed choices.
Vol. 434: Artificial Intelligence examines the ethical and legal questions surrounding AI, including how the technology may be used in the future.
Vol. 459: Digital World explores the internet's evolution and its role in everyday life, including algorithms and online safety, helping students navigate the digital age with greater confidence and critical awareness.
All titles are available in print and digitally through an Issues Online subscription. To find out more, email issues@independence.co.uk.