Teachers need AI literacy, ethics and agency before classrooms can scale AI

A new review published in Education Sciences finds that AI can support more flexible and responsive teacher preparation, but only when tied to pedagogy, ethics, equity and professional judgment.

The study, titled Sustainable AI Integration in Teacher Education: From Personalised Learning to Signature Pedagogies, reviews literature from 2019 to 2025 on AI in teacher education and offers a post hoc conceptual synthesis of eight recurring dimensions of sustainable AI integration, covering AI-driven personalization, pedagogical enhancement, professional development, ethical use, collaboration, infrastructure, AI literacy and governance.

AI is shifting teacher education from tool use to pedagogical design

AI is now increasingly embedded in curriculum design, practicum work, field experience, teacher mentoring, and digital literacy - a shift that changes the role of teachers and teacher educators. They are no longer expected only to adopt new systems, but to evaluate, adapt, and question them.

For pre-service teachers, the review suggests that AI can create safer environments to test instructional strategies before entering full classroom responsibility. Simulations, adaptive learning modules and AI-supported feedback can help novice teachers rehearse classroom decisions, reflect on mistakes and refine their professional identity. For in-service teachers, AI is more often linked to professional development, classroom management, ongoing mentoring and data-informed improvement.

This does not mean AI should replace the teacher's role. The reviewed literature consistently presents AI as a scaffold for professional judgment, not as a substitute for it. Teachers still provide empathy, moral responsibility, cultural awareness, classroom relationships and context-sensitive decisions that AI systems cannot supply on their own.

The review primarily focuses on "signature pedagogies" - teaching approaches that become distinctive to a teacher's professional identity and practice. The authors state that AI may help teachers develop these approaches by giving them more ways to design, test, revise and personalize instruction. For example, AI may support inquiry-based or project-based learning by helping teachers build lesson sequences, generate formative assessments, create simulations or adapt materials to student needs.

AI-supported teacher education is strongest when teachers remain active designers. Rather than accepting AI-generated content as ready-made instruction, teachers need to align it with learning goals, student context, curriculum requirements and ethical responsibilities. AI can expand instructional possibilities, but professional judgment determines whether those possibilities become good teaching.

While much of the debate around AI in education remains focused on efficiency, the paper argues that sustainable integration cannot be measured only by whether AI saves time or automates tasks. It must also be judged by whether it strengthens reflective practice, improves pedagogical decision-making and supports teachers' long-term professional capacity.

The UAE context gives the review added policy relevance. The authors discuss the UAE's national focus on AI, including its National Artificial Intelligence Strategy 2031 and Vision 2071, and note that AI is set to become a core school subject from kindergarten through grade 12. In that setting, teacher education becomes a critical point of implementation. If teachers are expected to prepare students for AI-rich classrooms, they must first be trained to use AI responsibly themselves.

The review also compares broader international experiences, including national efforts to build AI awareness and digital literacy. It finds that teacher education systems cannot treat AI adoption as a technical upgrade alone. AI changes what future teachers need to know, how they plan lessons, how they assess learning and how they understand their responsibilities in digital classrooms.

Readiness gaps, bias and privacy concerns remain barriers

Teacher readiness was found to be uneven. Across the literature, adoption is shaped by perceived usefulness, ease of use, AI confidence, self-efficacy, digital competence and institutional support. Teachers are more likely to adopt AI when they believe it improves teaching, reduces workload and supports student learning. But enthusiasm weakens when training is limited, infrastructure is poor or ethical guidance is unclear.

Several reviewed studies point to gaps in teacher confidence. Some teachers use generative AI tools such as ChatGPT, Canva or Claude AI for planning, writing, feedback or classroom preparation, but many remain unsure how to integrate them responsibly into teaching. Others lack formal AI training or hold misconceptions about what AI systems can do. The review notes that some teachers even associate AI with human-like consciousness, showing the need for deeper AI literacy.

AI literacy, according to the paper, is not just technical skill. It includes understanding how AI systems work, recognizing their limits, identifying bias, protecting data, evaluating outputs and knowing when human judgment should override automation. The authors argue that teacher education programs must prepare future teachers to ask not only how to use AI, but why, when and under what conditions it should be used.

Data privacy is a major concern because AI-enabled education systems often rely on student and teacher data. If governance is weak, sensitive information may be exposed, misused or fed into systems without clear accountability. The paper also highlights algorithmic bias, especially when AI tools are trained on datasets that reflect unequal social, linguistic or cultural assumptions.

The review raises concerns about Western-centric knowledge in AI-generated educational content. If teacher education programs rely heavily on commercial systems trained on narrow knowledge bases, they may reproduce cultural bias, marginalize local knowledge and weaken contextual relevance. This is especially important in multicultural systems such as the UAE, where schools operate across diverse curricula, languages and cultural settings.

Academic integrity is another concern. Generative AI can help teachers plan and create materials, but it can also encourage plagiarism, shallow learning or overreliance on automated text. The authors argue that teacher education must prepare teachers to design assessments and learning tasks that preserve critical thinking, originality and evidence-based reasoning.

The digital divide also threatens sustainable adoption. Access to AI tools, high-speed internet, devices, subscriptions and training varies across regions and institutions. If AI integration is driven by expensive commercial tools, better-resourced schools and teachers may gain advantages while under-resourced communities fall further behind. The review warns that AI adoption without equity planning may deepen the very inequalities it claims to solve.

Another recurring issue identified in the study is teacher's professional agency. Some studies reviewed by the authors suggest that AI can expand teacher leadership by freeing time for mentoring, planning and data-informed decision-making. Others warn that AI can reduce autonomy if teachers are pressured to follow automated recommendations or institutional mandates without room for professional interpretation.

Sustainable AI in teacher education depends on keeping teachers in control of pedagogy. AI should support teacher expertise, not narrow it. When teachers are trained to critically evaluate AI outputs, they can use the tools to strengthen instruction. When they are not, AI can become a source of dependence, confusion or compliance.

Sustainable AI requires policy, infrastructure and professional capacity

Sustainable AI integration is a system issue - it depends less on the sophistication of any single tool and more on how teacher education aligns pedagogy, professional development, ethics, infrastructure and governance.

The authors synthesize eight recurring dimensions from the literature.

1. AI-driven personalization refers to using performance data and adaptive systems to tailor teacher training.
2. Pedagogical enhancement refers to using AI to support inquiry-based, problem-based and project-based learning.
3. Continuous professional development covers AI mentors, feedback systems, simulations and ongoing learning opportunities.
4. Ethical and inclusive AI use addresses data privacy, bias reduction and responsible implementation.
5. Collaborative learning tools support peer discussion, curriculum co-design and shared reflection.
6. Scalable infrastructure ensures access across different settings, including remote and under-resourced schools.
7. AI literacy and teacher identity connect technical knowledge with professional values.
8. Policy and governance frameworks provide the rules, safeguards and institutional readiness needed for responsible use.

A teacher may have access to an advanced tool, but without training, ethical guidance and institutional support, that access may not lead to better teaching. Policymakers need to set clear rules on data privacy, algorithmic fairness, accountability and access. National AI strategies in education must be matched by teacher preparation, curriculum redesign and practical guidance for schools. Ambitious policy will not translate into classroom value unless teachers know how to apply AI in ways that protect students and serve learning goals.

AI literacy should be embedded in teacher preparation, not treated as a short add-on workshop. Pre-service teachers need opportunities to test AI tools, evaluate their outputs, design AI-supported lessons, discuss ethical dilemmas and reflect on how AI affects their professional identity.

The review supports continuous professional development for in-service teachers rather than one-time training. AI tools change quickly, and teachers need recurring support to adapt. Professional learning should include technical practice, pedagogical design, ethical analysis and opportunities for teachers to share examples from real classrooms.

It further points out the importance of hybrid teaching models that combine human instruction with AI support. This approach avoids two weak extremes: rejecting AI entirely or handing instructional authority to automated systems. In a sustainable model, teachers use AI to expand planning, feedback and personalization while preserving the human relationships and judgment at the center of education.

It's important to note that the authors acknowledge a few limitations in their work. The review is based on selected literature rather than new classroom data, and its conceptual framework is a post hoc synthesis rather than an empirically validated model. The authors also note that much of the available research still lacks long-term evidence on how AI affects teacher quality, student outcomes, professional identity and metacognitive skills over time.

More longitudinal research is needed to understand whether AI-supported teacher preparation produces lasting improvements or merely short-term efficiencies. Comparative studies are also needed across different cultural, institutional and policy contexts, rather than assuming that one model fits all education systems.