Artificial intelligence is no longer a future issue in children’s lives. It is already being used for homework, information, translation, creativity and, in some cases, advice about personal worries. A new issue brief published by the UNICEF Office of Strategy and Evidence – Innocenti in June 2026 provides an important snapshot of how children and their parents or caregivers are using and understanding AI. The findings show that children are adopting AI rapidly, but not equally—and that adult knowledge, public policy, research and safety systems are struggling to keep pace. The brief draws on nationally representative household surveys conducted as part of the second phase of the Disrupting Harm project, led by UNICEF Innocenti, ECPAT International and INTERPOL, with funding from Safe Online. Approximately 1,000 internet-using children aged 12–17 and 1,000 parents or caregivers were surveyed in each of ten countries: Armenia, Brazil, Colombia, the Dominican Republic, Jordan, Mexico, Montenegro, North Macedonia, Pakistan and Serbia.

AI Has Already Entered Children’s Everyday Lives

Across the ten countries, between 18 and 50 per cent of surveyed children reported having used AI at least once. Together, this represents an estimated minimum of 20 million children. The figures should not be interpreted as a universal global rate. AI use varied significantly between countries, and the surveys focused on children who already had internet access. Nevertheless, the findings confirm that AI is becoming part of ordinary childhood rather than remaining a specialist or experimental technology. Older adolescents were generally more likely to have used AI than younger children. Across the countries surveyed, children aged 15–17 were between 5 and 18 percentage points more likely to use AI than those aged 12–14. This matters because children are not simply encountering AI through one dedicated application. AI is increasingly embedded in search engines, social media platforms, educational tools, image-generation services and digital communication systems. For many children, it may therefore become difficult to distinguish between “using the internet” and “using AI”.

Homework Is the Most Common Use—but Learning Requires More Than an Answer

Among children who had used AI:

• 60 per cent used it to help with homework;
• 47 per cent used it to find information;
• 30 per cent used it to create or edit images and videos;
• 24 per cent used it for translation;
• 11 per cent used it to seek advice about things that worried them.

The educational potential is clear. AI can explain difficult material, support translation, improve accessibility and help children explore new ideas. For children who lack access to private tutoring or specialist educational support, it may also create new learning opportunities. But using AI for homework is not automatically the same as learning. UNICEF highlights the risk of cognitive offloading: allowing AI to perform the thinking, reasoning or problem-solving that children need to practise themselves. When children receive complete answers without guidance, they may produce finished work while missing the process through which learning actually takes place. The challenge is therefore not to exclude AI from education. It is to help children use it as a tool that supports thinking rather than replaces it. Children need to learn how to question AI-generated information, recognise errors and bias, verify sources, compare perspectives and understand that a confident answer is not necessarily a correct one.

A Growing AI Divide Within Families

One of the most important findings concerns the gap between children and adults. Across the ten countries, an average of 38 per cent of children had used AI, compared with only 12 per cent of parents or caregivers. At least three in four parents or caregivers had never used it in most of the surveyed countries. Children were therefore more than three times as likely as their parents or caregivers to be AI users. This intergenerational divide can create a serious protection gap. Adults may be expected to guide children through technologies they have never used, do not fully understand and may not even recognise. This is not a reason to blame parents. The speed of technological development means that many adults have had little opportunity to develop AI literacy. Parents, caregivers and teachers need accessible information, practical training and support that allows them to learn alongside children. Effective parental guidance cannot depend only on prohibition. The report indicates that excessive restriction may reduce children’s opportunities to explore new tools and develop digital skills. Protection should therefore combine age-appropriate boundaries with communication, shared exploration and critical understanding.

When Children Turn to AI for Emotional Advice

The finding that 11 per cent of child AI users sought advice about things that worried them deserves particular attention. In one surveyed country, the proportion reached 21 per cent. AI may feel accessible, private and non-judgemental. A child may find it easier to type a question into a chatbot than to speak to a parent, teacher, psychologist or other trusted adult. In some situations, AI may provide basic information or encourage a child to seek help. But AI systems are not human relationships, mental health professionals or safeguarding services. They may misunderstand context, provide inappropriate advice or fail to recognise that a child is in danger. Repeated use of AI companions may also contribute to emotional dependency, especially when systems are designed to appear caring, responsive or human-like. Children must know that AI can simulate empathy without understanding their experience. They also need clear and safe pathways to human support when they are distressed, frightened or facing abuse. AI literacy must therefore include emotional literacy: understanding what these systems are, what they are not and when a human being needs to become involved.

Children Recognise Some Risks—but Awareness Is Uneven

The children surveyed expressed concern about several potential harms:

• 34 per cent were concerned about AI being used to scam or deceive people;
• 32 per cent were concerned about AI spreading false information;
• 26 per cent were concerned about fake sexual images or videos;
• 15 per cent were concerned about receiving bad advice from AI;
• 14 per cent were concerned that AI might affect their future employment.

However, nearly one quarter reported none of these concerns, while another 23 per cent said they did not know or preferred not to answer. Awareness differed considerably across countries. In some, only around one in ten children expressed no concern. In others, approximately three in ten children did not identify any of the risks presented to them. Digital confidence alone does not guarantee safety awareness. The report found that children with stronger digital skills were more likely to use AI, but they were not necessarily more concerned about its risks. Knowing how to operate a tool is therefore different from understanding its consequences. AI literacy must address not only technical competence but also misinformation, privacy, manipulation, bias, emotional dependency, harmful content and the commercial interests behind digital systems.

Synthetic Images Can Produce Real Abuse

The most serious finding concerns the use of AI in child sexual abuse and exploitation. AI tools can manipulate an ordinary photograph of a child to create a sexually explicit deepfake. One method, often described as “nudification”, digitally alters clothing to produce a fake nude image. UNICEF estimates that, across 11 surveyed countries, images of at least 1.2 million children were manipulated into sexually explicit deepfakes within a single year. In some countries, this was equivalent to approximately one in every 25 children—or one child in a typical classroom. The fact that an image is artificially generated does not make the harm artificial. The child’s identity and appearance are being used without consent. The image may be distributed among peers, posted online or used for bullying, humiliation, coercion or sexual extortion. It can damage a child’s dignity, reputation, relationships, mental health and sense of safety. The report calls for stronger laws and governance frameworks, including the criminalisation of AI-generated child sexual abuse material, safety testing before powerful models are released and greater accountability for platforms and technology companies. Responsibility cannot be placed on children to protect themselves from systems that were not designed with their rights and safety at the centre.

The AI Divide Is Also a Rights Divide

The report identifies another challenge: unequal access. AI usage among children in the surveyed countries was lower than levels reported in several higher-income settings. For children without meaningful internet connectivity, suitable devices or opportunities to develop digital skills, the AI revolution may create a new layer of inequality. Children with greater access may benefit from new educational, creative and informational tools. Children who are excluded may fall further behind in societies increasingly shaped by automated systems. Closing the AI divide therefore does not simply mean giving every child access to technology. It means providing meaningful connectivity, inclusive design, accessible education, appropriate safeguards and the skills needed to use AI critically and safely. Access without protection is inadequate. Protection without access can also deepen inequality.

What Needs to Happen Now

Based on its findings, UNICEF identifies five central priorities:

1. Close the AI divide within and between countries by investing in meaningful connectivity and digital infrastructure, particularly for marginalised children and communities.
2. Invest in child-centred research on the effects of AI on children’s learning, development, relationships, mental health and well-being.
3. Build AI literacy among children, parents, caregivers and educators so that families can explore opportunities and manage risks together.
4. Create safer AI systems for children, with strong privacy protections, transparency and safety settings enabled by default.
5. Protect children from AI-enabled sexual abuse and exploitation through effective legislation, safety testing, corporate accountability and platform responsibility.

These priorities reflect a fundamental principle: children should not be required to adapt to unsafe technologies. Technologies must be designed and governed in ways that respect children’s rights.

Why This Matters for Connect to Protect

The findings of the UNICEF Innocenti brief speak directly to the central philosophy of Connect to Protect – Children & Youth at Risk: Childhood has changed. Child protection must change too. Children now live across interconnected physical and digital environments. Their education, friendships, creativity, emotional expression and exposure to harm can no longer be understood as belonging exclusively to either an “online” or an “offline” world. AI brings opportunities, but it also reshapes familiar child-protection challenges. Educational inequality, misinformation, sexual exploitation, emotional vulnerability and family support are not new issues. What is new is the speed, scale and technological form through which they can emerge. This is why AI cannot be treated as a purely technical subject. It is also a child-rights, education, mental-health, safeguarding, justice and public-policy issue. The discussion at Connect to Protect will bring these perspectives together, helping researchers, professionals, policymakers, international organisations and technology experts move beyond both uncritical enthusiasm and fear-driven responses. The question is no longer whether AI will become part of childhood. It already is. The question is whether the systems responsible for protecting children will evolve quickly enough to respond.

Source

UNICEF Office of Strategy and Evidence – Innocenti. (2026). Snapshot of AI Usage and Concerns Among Children and Parents: Insights from 10 Countries. UNICEF Innocenti, Florence, June 2026. https://www.unicef.org/innocenti/media/13266/file/UNICEF-Innocenti-AI-Brief-2026.pdf