The generative AI revolution is reshaping teacher education. Moving beyond speculative debate, AI is now increasingly present within classrooms, lesson planning, staffroom conversations, and the everyday practices of schools and universities.

Within our Primary PGCE programme, early discussions among ITE colleagues reflected familiar sector-wide concerns. Should AI use be restricted or even banned? Would trainees stop thinking for themselves? Would lesson planning become a case of copying and pasting AI-generated content without understanding the underlying pedagogy or subject knowledge?

Alongside these questions were equally strong concerns from school leaders and mentors. Some framed AI as a form of “cheating” or an unfair shortcut into the profession. Others argued that trainees should first “earn their stripes” in planning and preparation before using tools seen as reducing workload. Underpinning much of this was a deeper concern: if thinking, planning and knowledge generation can be outsourced, what does this mean for professional integrity, credibility, creativity and the complex craft of teaching itself?

These concerns were entirely recognisable across the sector. However, what quickly became clear was that our trainees were already using generative AI before they entered the programme. The question was therefore no longer whether AI should enter ITE, but how we in ITE should respond.

This prompted a shift in our stance: from gatekeepers to guides. Our role became focused on supporting trainees to engage with AI critically, ethically and professionally within the context of education, while continuing to develop the intellectual and pedagogical foundations of teaching. This aligns with UCET’s Intellectual Base of Teacher Education (IBTE), particularly the development of teachers as reflective, research-informed and epistemically grounded professionals.

To explore this further, we established a collaborative research team and undertook a longitudinal case study within our Primary PGCE programme. TeachMate AI was selected as our primary AI platform due to its closed-model design and explicit alignment with the English National Curriculum. Unlike open-access AI platforms drawing from the wider internet, TeachMate AI offered a more controlled environment, reducing risks relating to data sourcing, inappropriate outputs and curriculum misalignment. Importantly, its use was not mandated; it was offered as one ‘tool’ available to trainees.

Working closely with trainee teachers and school partners, we explored how AI was being used and understood in practice within school placement settings. Rather than treating AI as a theoretical issue, we focused on lived experience within classrooms and placements.

Across the year, we gathered perspectives from:

• Primary trainee teachers
• School-based mentors
• School leaders
• University academics

These conversations provided a grounded and often candid picture of emerging practice, tensions and opportunities across the sector.

Initial analysis and findings

Our analysis identified key perceptions, practices and tensions regarding AI use in ITE. Importantly, many of the assumptions held at the outset of the project were not borne out in practice.

AI was most commonly used to support workload-intensive tasks such as generating model texts, scaffolds, resources and adapted materials, rather than replacing lesson planning. Crucially, trainees did not accept outputs uncritically. They evaluated, adapted and refined materials in response to pupil need, curriculum expectations, accessibility and classroom context. Professional judgement and pedagogical reasoning remained central to decision-making.

Interestingly, trainees often demonstrated greater critical rigour when engaging with AI-generated materials than when using commercially produced schemes of work adopted by schools. Several described published schemes as appearing “fixed” or inherently “correct”, leading to less interrogation or adaptation. AI outputs, by contrast, seemed to prompt more questioning, refinement and scrutiny of suitability.

This raises a wider sector issue. Concerns about reduced creativity and teacher autonomy are not unique to AI; they reflect longer-standing debates about heavily structured schemes and curriculum packages already embedded in many schools. In many contexts, trainees appeared to apply greater criticality and further adaptation to AI outputs than to established, school-led materials.

At the same time, there was significant variation across schools in AI awareness, policy and practice. This created a fragmented landscape for trainees navigating different expectations, sometimes resulting in informal or “hidden” AI use within planning processes.

From findings to sector debate

Taken together, these findings challenge our early assumptions that trainees would simply outsource professional thinking to AI. Instead, AI was primarily used to reduce workload while maintaining clear ownership of pedagogical decision-making. Rather than diminishing professional identity, it often appeared to support emerging teacher autonomy and reinforce critical engagement with practice.

This feeds into a current sector debate: whether AI represents a threat to teacher expertise or a tool that can be integrated in ways that strengthen it. Our evidence and study so far suggest the latter, but only where AI use is framed within a knowledge-rich model of teacher education rather than a tool-driven one. Our findings strongly align with UCET’s Intellectual Base of Teacher Education (IBTE), particularly the development of teachers as epistemic agents: professionals who critically evaluate knowledge, exercise informed judgement, and adapt practice to context rather than reproducing resources uncritically.

Crucially, our data also highlights a parallel challenge for the sector: maintaining epistemic agency requires resisting both AI dependence and scheme dependence. In other words, quality teacher education must ensure trainees are not positioned as consumers of pre-generated content, whether from algorithms or commercial materials, but as active constructors and interrogators of professional knowledge. Notably, this level of critical engagement with both AI and curriculum materials was not widely anticipated by ITE colleagues or school leaders at the outset of the project.

Year one

Year one confirms that responsible AI use is already emerging within ITE. However, it also reinforces the need for sustained attention to how AI intersects with teacher knowledge, identity and professional judgement.

Key implications for trainee development include:

• Contextual awareness: AI outputs must be adapted to pupil need, curriculum intent and classroom context.
• Workload management: AI may support efficiency but requires time for structured review and quality assurance.
• Professional judgement: Teachers must retain responsibility, agency and accountability for AI-supported outputs.
• Subject and pedagogical knowledge: A strong knowledge base is essential to evaluate and verify AI content.
• Prompting as pedagogy: Effective use of AI depends on clear articulation of learning purpose, curriculum intent and teaching rationale.

Importantly, our findings reaffirm that AI cannot replace teachers or the professional knowledge base of teaching. Instead, the challenge for ITE is not whether AI should be used, but how it can be shaped to strengthen professional knowledge, judgement and pedagogical integrity within a knowledge-rich model of teacher education.

Our evidence suggests that when engaged with critically and deliberately, AI can support rather than erode epistemic agency.

Looking ahead

As we move into the next phase of the project, we will continue to explore how AI is reshaping both school practice and ITE provision. Perhaps the most relevant question for the sector is this:

How do we prepare future teachers to engage critically and confidently with AI in ways that enhance pupil learning, while ensuring teacher education remains knowledge-rich, professionally grounded, and committed to sustaining epistemic agency in an AI-shaped world?

Not a simple question—but one the ITE sector cannot afford to avoid.

Blog Authors: Karen Kilkenny and Natalie Jayson, on behalf of the research team at the University of Manchester, Liz Birchinal, David James and Andrew Rhodes.