A report for UNESCO proposes replacing “at what age?” with “for what purpose and on what evidence?” when using AI in the classroom
Sonia Livingstone and three co-authors argue that there is no single right age for using AI in education and call for decisions based on child development and clear educational aims.
“Is there a right age for AI in education? Deciding whether AI is age-appropriate and why” is a research report by Sonia Livingstone, Kim R. Sylwander, Sandra El Gemayel and Ayça Atabey, published by UNESCO in 2026. Livingstone, professor of Social Psychology in the Department of Media and Communications at the London School of Economics, presented it at UNESCO’s Digital Learning Week and published a summary on the Media@LSE blog on 22 September 2026. According to that summary, the text was prepared to support policy briefs on the governance of AI in education. It should be noted that this article relies on the author’s summary and on the library entry, because the full text could not be opened in UNESCO’s repository.
The thesis is that there is no single right age, because child development varies and depends on context. Hence the proposal to change the question “at what age?” to “for what purpose, under what conditions and on what evidence?”, a formulation also found in the library entry. Livingstone calls for decisions to start from child development and clear educational aims; without them, she warns, education ministries risk making decisions without foundation. She also distinguishes between teaching about AI, that is, AI literacy, which she considers a priority from an early age, and using AI to teach, an area in which she argues there is little evidence. As a starting point in the face of that uncertainty, she writes that “the precautionary principle is a reasonable starting point”.
The summary offers a few data points. The UN’s Independent International Scientific Panel on AI looked for evidence on the effects of AI on society, including education, and found little on the benefits and more on the risks. In the Rights.AI research, carried out in Kenya, Brazil, India and Thailand, children used AI, often unsupervised, to find information, create, have fun and ask for advice. In the United States, 73% of teenagers said they had never talked to a teacher about what AI is and how it works. The author also mentions privacy risks and unannounced changes to the tools schools use, and notes that there are age restrictions in places such as Norway, New York and California.
In the library, the document sits alongside others that approach children and AI from different angles. The World Economic Forum report on child safety and well-being (our article) agrees that age alone is not enough as a criterion, although it is aimed mainly at those who design and deploy systems. According to the entry for the message itself, the text was presented at UNESCO’s Digital Learning Week, where other documents were published, among them “The Algorithm in the Room” (in the library).
For teachers, the value lies in the change of question rather than in a rule: before introducing a tool, it is worth knowing what educational purpose it serves, what evidence supports it and whether it is a matter of teaching about AI or teaching with it. The limit is that, by its own thesis, it offers no threshold that can simply be applied in a school’s rules. As a first read, the Media@LSE summary is short and links to the full report.
A report for UNESCO proposes replacing “at what age?” with “for what purpose and on what evidence?” when using AI in the classroom
Sonia Livingstone and three co-authors argue that there is no single right age for using AI in education and call for decisions based on child development and clear educational aims.
“Is there a right age for AI in education? Deciding whether AI is age-appropriate and why” is a research report by Sonia Livingstone, Kim R. Sylwander, Sandra El Gemayel and Ayça Atabey, published by UNESCO in 2026. Livingstone, professor of Social Psychology in the Department of Media and Communications at the London School of Economics, presented it at UNESCO’s Digital Learning Week and published a summary on the Media@LSE blog on 22 September 2026. According to that summary, the text was prepared to support policy briefs on the governance of AI in education. It should be noted that this article relies on the author’s summary and on the library entry, because the full text could not be opened in UNESCO’s repository.
The thesis is that there is no single right age, because child development varies and depends on context. Hence the proposal to change the question “at what age?” to “for what purpose, under what conditions and on what evidence?”, a formulation also found in the library entry. Livingstone calls for decisions to start from child development and clear educational aims; without them, she warns, education ministries risk making decisions without foundation. She also distinguishes between teaching about AI, that is, AI literacy, which she considers a priority from an early age, and using AI to teach, an area in which she argues there is little evidence. As a starting point in the face of that uncertainty, she writes that “the precautionary principle is a reasonable starting point”.
The summary offers a few data points. The UN’s Independent International Scientific Panel on AI looked for evidence on the effects of AI on society, including education, and found little on the benefits and more on the risks. In the Rights.AI research, carried out in Kenya, Brazil, India and Thailand, children used AI, often unsupervised, to find information, create, have fun and ask for advice. In the United States, 73% of teenagers said they had never talked to a teacher about what AI is and how it works. The author also mentions privacy risks and unannounced changes to the tools schools use, and notes that there are age restrictions in places such as Norway, New York and California.
In the library, the document sits alongside others that approach children and AI from different angles. The World Economic Forum report on child safety and well-being (our article) agrees that age alone is not enough as a criterion, although it is aimed mainly at those who design and deploy systems. According to the entry for the message itself, the text was presented at UNESCO’s Digital Learning Week, where other documents were published, among them “The Algorithm in the Room” (in the library).
For teachers, the value lies in the change of question rather than in a rule: before introducing a tool, it is worth knowing what educational purpose it serves, what evidence supports it and whether it is a matter of teaching about AI or teaching with it. The limit is that, by its own thesis, it offers no threshold that can simply be applied in a school’s rules. As a first read, the Media@LSE summary is short and links to the full report.
