Written evidence from the Royal Society (AIE0215)

Education committee

The Use of Artificial Intelligence and EdTech in Education

About the Royal Society

The Royal Society is the UK’s national academy of science and a Fellowship of the world’s leading scientists. The Royal Society’s fundamental purpose, as set out in its founding Charters of the 1660s, is to recognise, promote and support excellence in science and to encourage the development and use of science for the benefit of humanity.

Education is the primary mechanism through which scientific literacy, critical thinking, and the capacity to evaluate evidence are developed across society, and through which the next generation of scientists is cultivated. As AI becomes increasingly embedded in how knowledge is produced, communicated, and acted upon, AI literacy has become an inseparable part of this process. The Society has therefore developed a substantial and active programme of work on AI in education. Central to this work is the Society's position that AI literacy is not only a technical competence but a civic and intellectual capability relevant to all young people, encompassing both how AI works and its implications for society.

This submission sets out the Society’s written evidence to the House of Commons Education Committee’s inquiry into The use of artificial intelligence and EdTech in education. It focuses on the following areas: teacher AI literacy and professional development; curriculum coherence and the Curriculum and Assessment Review; AI literacy as a cross-curricular responsibility; and equity.

Executive Summary

•          AI is reshaping how young people learn, how teachers work, and how knowledge is produced and evaluated. AI may offer opportunities to support pupil engagement, personalise learning, reduce teachers’ administrative workload, and widen access to high-quality education

•          England's education system is adopting these technologies rapidly, but the literacy, governance, and infrastructure needed to do so safely and equitably are not yet in place.

•          Teacher AI literacy is a critical weak point. Only 14% of teachers are confident across all dimensions of AI literacy and only 18% have received any structured professional development. Teachers in schools with formal continued professional development (CPD) are roughly three times more likely to be confident than those without it, yet initial teacher training frameworks have not been updated to reflect AI literacy as a core professional competence.

•          AI literacy is barely present in the curriculum. Only 2% of teachers report it being covered across most subjects, and 37% say it is not addressed in teaching at all.

•          Without a consistent national approach to AI literacy, policy defaults to tool adoption: over-indexing on technical and practical dimensions at the expense of the ethical, civic, and societal understanding that all young people require. This risks repeating the failures of earlier digital and coding literacy initiatives, where fragmentation produced uneven provision rather than systemic change.

•          AI in education risks deepening existing inequalities if provision continues to be distributed along existing lines of advantage. The gender gap in teacher confidence, the primary/secondary disparity in CPD access, the infrastructure shortfall in rural and deprived schools, and the concentration of computing in settings that already serve advantaged pupils all require deliberate policy attention.

•          The evidence base on AI's effects on learning and cognitive development is limited. Some early findings show positive effects on motivation and engagement as well as some negative effects on pedagogy and cognitive dependency which will require further verification. Large-scale procurement decisions should not outpace the generation of robust evidence.

•          Addressing these challenges requires a long-term commitment: a cross-curricular framework with accountability built into inspection, a properly resourced teacher workforce, and infrastructure investment targeted at the schools most at risk of being left behind.

The impact on teaching

Teacher AI literacy and CPD

Teachers play a central role in how young people encounter and make sense of AI in formal educational settings, providing essential support and modelling of responsible, critical, and reflective use.[1] Where teachers are well-equipped, they are well-placed to model AI as a tool that amplifies rather than replaces human capability; supporting pupils to use it productively while maintaining the judgement to evaluate and question its outputs.[2] Teacher AI literacy must therefore be treated as a precondition for AI’s use in classrooms.

The existing evidence base gives cause for concern. A Royal Society weighted survey of over 9,000 teachers in England conducted in March 2026 found that only 14% are confident across all three dimensions of AI literacy (technical understanding, practical use, and the human and ethical dimension).[3] A third are confident in practical tool use only, and a third report limited confidence across all aspects. This is consistent with the Royal Society’s rapid review of AI literacy frameworks,[4] which found that awareness of AI tools does not imply the capacity to teach about AI critically. The review identified that most current AI-related CPD for teachers in England is informal, fragmented, and heavily skewed towards practical tool adoption rather than the critical, ethical, and socially contextualised understanding that effective AI literacy requires.

The data on provision reinforces this fragmentation. Only 18% of teachers report having received structured professional development on AI literacy, only 15% have had clear whole-school guidance, with most teachers relying on informal guidance (32%), personal initiative alone (18%), or no support of any kind (23%).[5] These gaps are not evenly distributed (detailed in Gaps and Disparities below).

Structural support does make a material difference. Teachers in schools with structured professional development and whole-school guidance are roughly three times more likely to be confident across all dimensions of AI literacy than those receiving none.[6] Yet clear policy and systems for building this capacity do not yet exist: initial teacher training frameworks have not been updated to reflect AI literacy as a core professional competence; three-quarters of school leaders identify lack of time and the cost of professional development as barriers; and 95% cite budget constraints as a barrier to broader technology uptake.[7]

The national curriculum and the Curriculum and Assessment Review

Based on a rapid review of AI literacy frameworks across the four UK nations and internationally, England does not yet have an appropriately well-defined framework for AI in education. Guidance has been issued but it does not provide a definition of AI literacy, specify what students should know or be able to do at different stages, assign cross-curricular responsibility, nor is it accompanied by the teacher development infrastructure or assessment reform that meaningful implementation would require. The Curriculum and Assessment Review represents a valuable but insufficient step in the right direction. The Society welcomes its recognition that digital literacy is a cross-curricular priority,[8] but is concerned that without further focused work, the ambition will not translate into consistent provision.

A large risk is that AI literacy becomes concentrated within computing, a subject already experiencing severe teacher shortages and low uptake among girls and students from less advantaged backgrounds[9] rather than being embedded across all subjects in a way that reflects AI’s relevance to every domain of human activity. The Royal Society recently convened expert educators and academics from across computing education and found a strong consensus that technical knowledge of AI must be situated alongside its social, ethical, and civic dimensions, and that this responsibility cannot be borne by computing teachers alone.[10]

Less than 2% of teachers say AI literacy is covered across most subjects. A further 11% say it is covered in some subjects, and 28% say it is addressed occasionally or informally. Critically, 37% say it is not currently addressed in teaching at all, and a further 20% are not sure, meaning that in roughly three in five schools, teachers cannot confirm that AI literacy is being taught.[11]

The growth of artificial intelligence highlights the need to rethink current education pathways so that all students gain a good understanding of how to apply the technology, alongside an awareness of how it shapes society, and the social and ethical challenges. Meeting that need requires young people to experience a broad education that covers STEM, humanities, languages, and vocational domains. Research commissioned by the Royal Society indicated that those who had studied the broadest range of subjects at age 16–18 demonstrated greater adaptability and career flexibility, with participants self-reporting cross-domain breadth as a protective factor against occupational obsolescence in a rapidly changing labour market.[12] The Royal Society intends to investigate these findings further using DfE longitudinal data sets in the near future. The Society’s upcoming Science for Society report further argues that early post-16 specialisation limits the adaptability of all young people, with the costs falling most heavily on those with fewest choices.[13] In the context of AI, a narrow curriculum risks producing young people who are technically confident with specific tools but lack the broader critical and evaluative capacities to navigate an AI-mediated world.

The impact on learning

Digital and AI literacy education

There are three distinct relationships between AI and education that policy must not conflate. Teaching and learning with AI refers to the use of AI tools to support learning and reduce teacher workload (the dominant focus of current DfE guidance and most industry-produced CPD). Teaching and learning for AI refers to preparing students for careers in computing and AI development (the focus of most specialist computing provision). Teaching and learning about AI - AI literacy- refers to equipping all young people with the knowledge, skills and values to use AI critically. It is this third relationship that current policy most consistently neglects, and that the Society's evidence identifies as the most important for all young people, regardless of their career aspirations.[14]

AI literacy itself spans three interdependent dimensions: a technological dimension (understanding how AI works), a practical dimension (using AI safely and responsibly), and a human dimension (critically engaging with AI's ethical, societal, and civic implications). A rounded AI-literate individual requires competence in all three. However, the human dimension consistently absent from current provision. Almost all the frameworks examined by the Society’s rapid review focus on the technological and practical dimensions.[15] Students might learn how to use AI tools but remain unprepared to interrogate questions of algorithmic bias, data privacy, democratic manipulation, or the labour conditions of AI development.

This is not an abstract risk. Across the four UK nations, provision is already diverging: England has issued broad guidance around AI adoption; Scotland has invested in civic and ethical dimensions at a societal level; Wales links AI loosely to responsible citizenship; Northern Ireland is still scoping approaches. None yet provides a coherent framework that embeds AI literacy across subjects with appropriate evaluation, resourcing, and inclusion built in. The risk is that this fragmentation entrenches a narrow, skills-only approach: repeating the pattern of earlier digital and coding literacy initiatives, where a proliferation of tools left teachers overwhelmed and provision uneven rather than scaled equitably.[16]

AI literacy must be a shared responsibility across all subjects, not concentrated in computing. History and citizenship already address questions of democracy, surveillance, and power; English and media studies already explore authorship, bias, and intellectual property; science and design technology cover data, sustainability, and AI in research. AI intensifies these topics in ways that make subject teachers across the curriculum natural and necessary partners in developing AI literacy. For AI literacy to be valued and delivered consistently, it also needs to be visible in school inspection frameworks; computing is currently underrepresented in Ofsted evaluations, and this structural invisibility weakens the incentive for schools to act.[17]

The Royal Society survey of teachers confirms gaps in the current policy approach. Only 2% of teachers report AI literacy being covered across most subjects in their school, and 37% say it is not addressed in teaching at all. These figures are worse in the schools where provision is most needed: primary schools, state-funded institutions, and those serving more deprived communities are all significantly less likely to report any coverage.[18] Without a clear national framework, defined standards, and cross-curricular accountability, this gap will widen rather than close.

AI’s effects on learning, cognition, and essential skills

The limited evidence on the effects of AI on learning outcomes is characterised by some positive findings on motivation and engagement. Well-designed AI tutoring tools have shown promising early learning gains in controlled settings, but mixed, or in some cases negative, findings on learning outcomes and cognitive development.[19] The pace of adoption in schools and education is currently outstripping the generation of reliable evidence.

This is an area where the evidence base is developing rapidly, and the Society urges the Government to invest in high-quality research on AI’s effects on learning and cognitive development across age groups and attainment levels, before making large-scale commitments to particular tools or platforms.

Gaps and disparities in access

Persistent and well-documented disparities in access exist across socioeconomic background, gender, geography, and SEND status. Computer science remains the least popular STEM subject among girls, making up just 23% of GCSE certificates,[20] and if AI literacy continues to be concentrated there, girls will be disproportionately excluded from meaningful provision.

The demographic gaps in teachers’ confidence are stark and should act as early indicators that these consistent and material challenges are likely to persist amongst the young people they teach without active policy change. 28% of male teachers report confidence across all dimensions, compared with only 10% of female colleagues. Teachers with SENCO responsibilities show the lowest confidence of any group in the survey, at 9%. Some 42% of teachers in the most deprived schools report that AI literacy is not addressed in teaching at all, compared with 33% in the most affluent.[21] Closing these confidence gaps is important for realising one of AI’s educational benefits in supporting more adaptive and inclusive learning.

Uneven investment in infrastructure continues to provide regional and institutional challenges. Evidence gathered for the Royal Society documents compounding inequalities: pupils from disadvantaged backgrounds are less likely to have reliable devices and connectivity at home, and their schools are less likely to have the infrastructure and staffing capacity to deliver high-quality digital education.[22] Without deliberate targeting of funding and support, AI in education will reproduce the digital divide that preceded it. In 2024/25, only 16% of secondary schools met the Goverment's connectivity and wiring technology standards. For primaries, it was 6%. Only 54% of rural schools had full fibre broadband.[23] Whilst connectivity has moderately improved, the Royal Society is concerned that without sustained and targeted investment, the AI divide will entrench the digital divide that preceded it.

Challenges and opportunities

The most significant challenges and risks

Three interconnected challenges stand out. First, a growing gap between the pace of AI adoption in schools and teachers’ capacity to engage with it critically. Adoption is proceeding informally and ahead of governance in the majority of schools, normalising uncritical AI use in ways that may be difficult to reverse. Second, widening inequalities in both AI tools and AI literacy: if provision continues to be distributed along existing lines of advantage (school type, geography, socioeconomic background, and gender), the technology will compound the attainment gap. Third, the risk that AI literacy prioritises tool confidence over genuine critical understanding - a risk amplified by the proliferation of industry-produced CPD resources that emphasise adoption over interrogation.[24]

If a shared framework, a prepared workforce and equitable infrastructure are in place, AI has the potential to enrich learning, support teachers, and extend access to high-quality education for young people across the UK. The Society hopes the committee will press Government to be ambitious: to produce a generation of young people who are not merely fluent consumers of AI but can evaluate, question, or shape it. The consequences for democratic citizenship, for the equitable distribution of opportunity, and for the UK's capacity to govern AI in the public interest extend well beyond the classroom.

May 2026