Only 14% of teachers confident across AI literacy, Royal Society finds

A survey of more than 9,200 teachers in England found limited training and patchy curriculum coverage, as the Royal Society argues that AI literacy should extend beyond computing lessons

AI in education research from the Royal Society finds only 14% of teachers are confident across AI literacy and just 18% have received CPD

Only 14% of teachers in England say they are confident across the full range of AI literacy skills, according to new Royal Society research that points to a widening gap between the spread of artificial intelligence and schools’ ability to teach students how to understand and question it.

The findings come from a weighted Teacher Tapp survey commissioned by the Royal Society and conducted on March 20, 2026, with more than 9,200 primary and secondary teachers responding.

Confidence was measured across three areas: understanding how AI works, using it safely and responsibly, and engaging critically with its ethical, societal and civic implications.

While 14% felt confident across all three, 33% were mainly confident in practical AI use and another 34% reported limited confidence across the board.

Training is also far from routine. Just 18% of teachers said they had received AI-related CPD or training courses, while 15% had clear whole-school guidance.

Nearly a quarter, 23%, reported receiving no support at all.

AI literacy remains largely outside the curriculum

The numbers become even starker when teachers were asked what students are actually being taught.

Fewer than 2% said AI literacy was covered across most subjects in their school. Another 11% reported coverage in some subjects and 28% said it appeared occasionally or informally.

Some 37% said AI literacy was not currently addressed in teaching at all, while a further 20% were unsure.

That leaves roughly three in five teachers unable to confirm that students in their school are receiving active AI literacy education.

The Royal Society wants AI literacy treated as a cross-curricular responsibility rather than left to computing departments. Its wider work categorizes AI education into three areas: teaching about AI, teaching for AI and teaching with AI.

Teaching about AI covers skills such as understanding how AI works, interpreting AI-generated content and considering ethical implications. Teaching for AI includes the mathematics, data and computing skills needed to build and manage AI systems, while teaching with AI focuses on educators using the technology in areas such as administration, learning and assessment.

The Society’s rapid review of existing AI literacy frameworks found that current approaches tend to place more weight on technical skills, including coding and algorithms, than on the wider societal and environmental aspects of the technology.

It also found no agreed UK-wide approach. England, Scotland, Wales and Northern Ireland are pursuing different strategies, but the Royal Society says none currently provides a coherent system-wide framework.

Training is closely linked with teacher confidence

The survey found a substantial difference between teachers receiving structured support and those left to work things out themselves.

Teachers in schools with professional development and whole-school guidance were around three times more likely to report confidence across all areas of AI literacy than teachers receiving no support.

For many teachers, however, support remains informal. Some 32% reported relying on informal guidance, while 18% depended on their own initiative.

Confidence also varied considerably between groups. Among teachers aged 20 to 29, 21% felt confident across all aspects of AI literacy, falling to 10% among those over 50.

There was an even larger gender gap. Some 28% of male teachers reported confidence across all three areas, compared with 10% of female teachers.

Secondary teachers were more likely to report confidence than primary teachers, at 18% compared with 11%.

Headteachers were below classroom teachers, with 9% reporting confidence across all dimensions against 14% of classroom teachers.

Teachers with SENCO responsibilities recorded the lowest figure of any group measured, at 9%. Just 12% of this group said they had received structured support.

School context also affected what students were being taught. In schools serving a higher proportion of students eligible for free school meals, 42% of teachers said AI literacy was not addressed at all. The equivalent figure was 33% in schools with a smaller proportion of students receiving free school meals.

The Royal Society warns that without a common baseline, access to AI literacy could increasingly depend on individual teachers or local initiatives.

Its work on AI in education will now continue with further evidence gathering intended to inform policy development, including how AI literacy could be defined and supported more consistently across the education system.