AI in education: a practical guide for schools and teachers

Three levels of AI in education

1. Teacher-facing AI helps educators plan lessons, create resources, draft assessments, differentiate materials and reflect on evidence of learning. The teacher reviews everything before it reaches students.
2. Student-facing AI includes tutors, feedback tools and chatbots that students use directly. It needs the most care around age, safeguarding and learning design.
3. System-level AI helps leaders spot patterns in attendance, assessment and workload to plan support.

Most schools find it easiest to begin with teacher-facing AI. Teachers stay in control, and the time saved goes back into teaching decisions.

The Royal Society offers a complementary lens that is useful for curriculum planning. It groups conversations about AI in education into teaching about AI (AI literacy: understanding and critically interpreting AI and its content), teaching for AI (the mathematics, data and computing skills needed to build and manage AI systems) and teaching with AI (using AI to reduce administrative load and personalise learning and assessment). A school strategy needs all three.

What the evidence says so far

Research on AI in education is moving fast, and much of it is early. The largest new source of evidence is PISA 2025, which found that the relationship between AI use and student performance is complex. Students who did not use AI for specific schoolwork tasks outperformed those who did. Frequent users who were taught to assess AI output tended to perform better than those who were not. The OECD also warns that AI literacy opportunities are more common among advantaged students. See what PISA 2025 says about students using AI.

International guidance points in the same direction. UNESCO describes its mandate as calling for a human-centred approach to AI, guided by inclusion and equity, and its 2026 collection “The algorithm in the room” asks who holds agency and oversight as AI enters classrooms. The OECD–European Commission AI literacy framework sets out what students should learn about AI.

Teachers are often ahead of policy. A rapid review commissioned by the Royal Society reports that over half of UK teachers already use AI for planning and marking, yet only 14% are confident across elements of AI literacy, with limited professional support. Access to tools is spreading faster than the knowledge to use them well.

Where AI helps teachers most

• Lesson planning: drafting a sequence around a learning objective, which the teacher then adapts to the class. See a connected AI workflow for the teaching week.
• Differentiation: creating support, core and extension versions of a task with the same learning goal. See differentiation and remediation.
• Assessment: generating question banks, hinge questions and exit tickets for formative assessment and retrieval practice.
• Feedback drafts: suggesting comments against a rubric that the teacher edits and personalises.
• Language support: preparing glossaries and parallel texts for multilingual classrooms.
• Accessibility: speech-to-text, transcripts and reading support for students who need them.

Risks to manage

A five-step adoption plan

1. Purpose: name one teaching need, such as reducing preparation time for differentiated practice.
2. Policy: agree who can use which tools, for what, and how outputs are reviewed. Start with our school AI policy guide.
3. Pilot: begin with a small group of teachers and a defined time frame.
4. Professional learning: give teachers time to practise with their own curriculum. The UNESCO teacher framework helps structure this.
5. Evaluate: gather feedback on quality, time and classroom fit before expanding.