Andreas Schleicher, Director of the Directorate for Education and Skills at the Organisation for Economic Cooperation and Development (OECD), delivered a keynote speech at the parallel session on "AI Education Development and Evaluation" during the 2026 World Digital Education Conference. His speech was titled "How to Make AI a ‘Scaffold’ Rather Than a ‘Crutch’: Global Trends". 

He noted that AI is not magic; rather, it acts as an accelerator and amplifier. It can accelerate the implementation of educational ideas, amplify the impact of educational practices, and offer us the opportunity to deliver excellent education to all people, benefiting everyone. At the same time, he cautioned that AI has two sides. It can empower teachers, but it can also reduce them to mere script‑readers. It can help overcome language barriers, yet it can also amplify biases and fuel segregation.

Focusing on generative AI (GenAI), he particularly observed that using general-purpose GenAI does not automatically improve learning outcomes and can sometimes hinder student understanding. For this reason, he argued that using GenAI in education requires certain conditions, making it necessary to distinguish between general-purpose GenAI and education-specific GenAI. General-purpose GenAI is not as capable as often assumed; only education-specific GenAI, designed for educational purposes and tested for safety, can realise greater value. He emphasised that while using GenAI for education, learning risks must be mitigated. Completing a task with GenAI does not mean learning through GenAI. As machines become smarter, human capabilities become more important. Students must learn to think first, and only then learn to ask questions. GenAI supported by teacher-led instructional and learning design can achieve better results. No GenAI-based lesson plan can replace a teacher’s professional judgment. Algorithms can provide suggestions, but the final decision lies with teachers, and teachers should not outsource assessment to AI.

He stressed the need to develop AI tools designed specifically for education, rather than repurposing general-purpose AI tools with minor modifications. For younger students, special attention must be paid to AI safety. We need global metrics and rigorous research to evaluate the use of AI in education. In this regard, the Global Digital Education Development Index (GDEI) launched by the China National Academy of Educational Sciences is a very important tool. To develop trustworthy AI for education, global cooperation is needed to enhance AI safety and harmonise standards.