What impact does AI have on children’s learning in schools?

https://doi.org/10.58248/RR109 

Artificial intelligence (AI) is increasingly used in schools to support teaching and learning. Current evidence suggests that AI may improve learner engagement, personalised feedback and aspects of literacy, particularly writing and language learning. However, the benefits depend heavily on how teachers integrate AI into teaching and learning.  

The evidence on longer-term effects on reasoning, creativity, critical thinking and neurological development remains limited. There are indications that AI may support pupils’ AI literacy and digital skills, but most studies are small-scale and short-term.  

AI could either reduce or widen educational inequalities, depending on access to technology, teacher capability and implementation.  

Better evidence in a UK context would support better understanding of the impacts of AI on children’s learning. 

Background 

AI-related teaching in schools has expanded rapidly since 2020 as age-appropriate AI teaching and learning tools have developed. However, there is a lack of robust empirical evidence on the effectiveness of AI-related teaching. Much of the recent literature comprises small-scale interventions and self-reported outcomes, with relatively little research undertaken in English schools. 

The UK Government set out its approach to generative artificial intelligence (AI) in education in 2025 and described the actions it is taking to “facilitate the use of the safe, responsible and effective use of generative AI in the education sector”. It noted that the evidence is still emerging on the benefits and risks of AI to teaching. 

The Commons Education Committee launched an inquiry examining the role of AI and EdTech in education in February 2026. 

Impact on student skills and development 

The most consistent evidence of the impact of AI on student skills relates to the use of generative tools, such as ChatGPT, to support feedback and personalised instruction. Studies have reported that using AI tools for feedback can improve some learning outcomes, including engagement during lessons and aspects of literacy. However, these effects vary considerably according to pupils’ age, subject, educational phase and the way AI is incorporated into teaching and learning.  

As with other forms of educational technology, international guidance emphasises the importance of teaching design, teaching expertise and curriculum context in determining how AI is used. AI therefore appears more likely to amplify established teaching practices, good and bad, rather than fundamentally alter them. 

Other studies have considered the effect of AI in education on pupils’ AI literacy. Some initiatives have been shown to improve pupils’ understanding of AI concepts, data, algorithms, ethical issues and the critical evaluation of AI-generated content, while also supporting aspects of computational thinking, such as breaking problems into smaller steps. However, the evidence remains relatively limited as it is derived largely from short-term studies. There is little indication that these gains are sustained over time or transferred consistently across different educational settings, school phases and subject areas. 

Evidence relating to wider developmental outcomes remains considerably less certain. To date, there is little empirical evidence examining whether AI use affects children’s neurological development. Similarly, research on executive function, critical thinking, creativity, oracy and broader cognitive development remains fragmented, with most studies relying on proxy measures collected over relatively short periods. While some evidence suggests that AI can encourage enquiry and dialogue when embedded within carefully designed teaching, these findings have yet to be confirmed through longitudinal or large-scale experimental research. 

Impact on cognition and independent thinking 

The widespread availability of generative AI has raised questions around how it may affect children’s cognition and independent thinking. Although it is much discussed within educational research, there is little empirical evidence on these effects for school-aged children. Existing school-based research has largely examined short-term outcomes, while evidence concerning longer-term cognitive effects is scarce and longitudinal studies are noticeably limited. 

From a developmental perspective, cognitive skills, including reasoning, self-regulation and problem-solving, are strengthened through repeated opportunities to engage with challenging tasks, supported by interaction with teachers and peers. As generative AI is able to perform some of these tasks, there have been concerns that learners may come to rely on AI tools for cognitive work (known as cognitive offloading), reducing opportunities to develop independent judgement and critical evaluation.  

Whether AI increases cognitive offloading in educational settings has yet to be established. However, studies suggest that where AI is used to support questioning, formative feedback, iterative drafting and classroom discussion, it can encourage reflection and deeper engagement with learning. These approaches may be more educationally valuable when teachers require pupils to evaluate or challenge AI-generated content rather than accept it uncritically. But there is insufficient evidence to determine definitively whether regular use of AI has lasting effects on reasoning, creativity or independent thinking. Nor is there robust evidence linking AI use with neurological development in children and young people.  

The principal gap in the evidence base is the absence of longitudinal research capable of distinguishing short-term improvements in task completion from enduring changes in cognitive development.  

Differences across subject areas 

Emerging evidence suggests that the uses and educational effects of generative AI differ across curriculum areas, although the evidence remains limited and uneven and current evidence does not allow robust comparisons of its effectiveness between subjects.  

Studies have examined its use in English and language learning, including for writing, revision and feedback, as well as in mathematics, science and computing. In mathematics and science, studies have explored uses including problem-solving, explanations, feedback and support for scientific inquiry. However, findings vary according to the task and the way AI is incorporated into teaching, and there is limited evidence about how it may affect pupils’ longer-term conceptual understanding and independent problem-solving.  

In the subject of computing, AI can be both a learning tool and a topic of study. A systematic review of 45 empirical studies of school-aged children found that teaching about AI can support pupils’ AI literacy, including their knowledge and understanding of AI systems and aspects of AI-related thinking. However, studies are commonly short-term and use relatively small samples, leaving limited evidence about whether these effects persist or transfer to other contexts. 

Current evidence therefore does not establish that AI is consistently more effective in particular curriculum areas. Its educational value appears to vary according to the nature of the learning task and the way teachers integrate AI into classroom practice. 

Inequalities associated with AI and EdTech 

The educational opportunities and risks associated with AI are unlikely to be experienced equally. Existing evidence suggests that AI could either reduce or reinforce educational inequalities, depending on how it is implemented in schools and the education system. As with other kinds of educational technology, outcomes are influenced by differences in regional and local digital infrastructure, teacher expertise, school resources and pupils’ social and economic circumstances. Pupils from disadvantaged households may therefore have fewer opportunities to access and benefit from AI-supported learning, potentially reinforcing existing attainment gaps if these differences are not addressed. 

AI also raises questions about fairness, transparency, trust and accountability. Generative AI systems are trained on large datasets that may reflect biases, which can influence the quality and accuracy of their outputs. Such biases may disproportionately affect groups of pupils who are under-represented in training data, or for whom AI systems perform less reliably. Current international guidance therefore emphasises that AI-generated content should remain subject to professional judgement and critical evaluation by teachers.  

For pupils with Special Educational Needs and Disabilities (SEND), AI presents both opportunities and risks. AI may improve accessibility for some pupils, including through personalised learning and assistive applications, and may support greater participation and independence. However, there is little evidence about these effects in school settings, particularly across different types of need. AI systems that are not designed with diverse learners in mind may also create new barriers or reproduce existing forms of exclusion. 

Whether AI tools are implemented equitably is likely to depend on how able teachers are to integrate it into their lessons. Schools differ considerably in the professional development and technical support available to teachers. International guidance emphasises that AI should be integrated within purposeful teaching and learning approaches rather than adopted as a technological solution to teaching and learning in its own right. Without sustained investment in professional learning and clear governance arrangements, there is a risk that differences in implementation between schools will contribute to existing educational inequalities rather than reduce them.  

Key questions for policymakers 

• What forms of longitudinal research are needed to understand the effects of sustained AI use on children’s learning and cognitive development?  
• How might the government balance the potential benefits of AI in education with its potential harms in the context of limited evidence and rapid technological change? 
• What evidence is needed to understand how AI use in schools affects existing educational inequalities in the UK? 
• What support is needed to ensure that pupils from disadvantaged backgrounds can access high-quality AI learning tools? 
• What role might AI play in supporting children with SEND, or who speak English as a second language? 
• How might the government ensure effective procurement of AI-related tools within the education sector?  

Acknowledgements 

Professor Sandra Leaton Gray is Professor of Education Futures at the UCL Institute of Education. 

Questions about this briefing should be referred to Oliver Bennett MBE, (post@parliament.uk), who acted as parliamentary lead for this work.