UNICEF’s child-centered AI guidance asks more than whether a tool is popular

The starting point

UNICEF’s December 2025 guidance sets out ten requirements for child-centered AI. The guidance highlights privacy, safety, inclusion, accountability and children’s development. It also identifies AI companions among the issues addressed in the update. This guidance is a framework for evaluation, not a certification of particular products.

Child-centered means asking whose experience defines success

A child’s ability to use an interface does not establish their ability to assess its motives, limits or consequences. Nor does an adult’s confidence remove the need to listen to children’s actual experience. A school or ministry must consider both protection and agency.

Questions about a chatbot should include how it responds when it is wrong, how easy it is to disengage and what information it encourages a young user to disclose. These are design and governance questions as well as questions about individual choices.

A product can be easy to operate without being appropriate for every child or purpose. An adult evaluating it may notice speed, convenience or an attractive interface. A child may encounter something different: confusion about who is speaking, difficulty leaving a conversation, or uncertainty about whether a response is true. These are possibilities to investigate, not harms measured by this article.

The value of a child-centered approach is that it does not allow the purchaser's convenience to become the only measure. It asks how the system relates to children's development, participation and protection. Those concerns should enter the decision before a tool is introduced, not only after someone raises a complaint.

UNICEF's guidance is a framework for that work. Its authority as a source does not convert every product using similar language into an approved service. A school or ministry still needs to examine the particular tool, the intended users and the circumstances of use. This article offers educational analysis, not an individual assessment of a child's needs.

Protection and participation should inform each other

Adults have responsibilities that should not be handed to children. A child should not have to understand a supplier's technical architecture or contractual language in order to receive appropriate protection. Leaders choosing a service need to inspect its information practices, controls and support arrangements themselves.

At the same time, excluding children's experience can leave adults unaware of how a tool is actually understood. A service may behave differently in ordinary use from the demonstration an adult saw. Age-appropriate listening can reveal misunderstanding or barriers without requiring young people to disclose private conversations in front of a group.

The two duties can reinforce one another. An institution can make protective decisions while offering a safe way for children to explain what they find useful or confusing. Participation does not mean making children responsible for approving the risks. Protection does not mean assuming that their account of the experience is irrelevant.

A religious interface can add a layer of authority

A tool placed on a church or faith-school website may appear to carry the institution's endorsement. If its language sounds pastoral or its visual design invokes a sacred figure, users may bring expectations that go beyond ordinary information retrieval. Whether particular children actually interpret it that way would require evidence; the possibility deserves examination in design review.

An institution should ask what the interface promises. Does it present itself as a source finder, a teacher, a friend or a spiritual guide? Can a young user explain that role in ordinary language? The answer should not depend on a lengthy disclaimer that contradicts the rest of the presentation.

A bounded tool that helps locate a public passage is a different proposal from a system inviting intimate spiritual disclosure. Treating both as “AI for young people” obscures the decision. Religious leaders should evaluate the intended relationship as well as the content of sample answers.

Test ordinary use, not only dramatic failures

A sensible evaluation includes mundane questions. Can the system admit that it lacks information? Does it identify a source the user can actually open? Does it encourage checking with an appropriate person when the question goes beyond its purpose? A service need not produce a spectacular error to be confusing or unhelpful.

Use fictional, non-sensitive scenarios for a demonstration. Do not enter a real child's private account merely to make a test realistic. Document the product version and settings so that the result is tied to what was actually examined. A later change may require another look.

Testing also has limits. A few examples cannot establish that a service is safe in every conversation. The institution needs a way to receive concerns and change or stop the use when evidence warrants it. A favorable initial demonstration should not become a permanent exemption from attention.

The case for helpful assistance should remain open

Children may benefit from accessible explanations, support in preparing questions or help with a clearly defined learning task. A responsible approach should examine those opportunities rather than assume that avoiding AI is the only form of care. The purpose and context matter.

For example, an educator might compare a student's understanding before and after a bounded activity, rather than count how many answers the tool produced. The evaluation could ask whether the student can explain the reasoning or locate the relevant source. Such an exercise is a proposed method, not a result demonstrated in this report.

An alternative should remain available when a learner cannot or does not wish to use a service. Otherwise the institution may confuse participation in education or community life with acceptance of one commercial interface. The practical form of that alternative will depend on the setting and resources.

Christian reflection should make adult responsibility more concrete

The account of Jesus welcoming children challenges a community to attend to people who can be overlooked in decisions made for institutional convenience. Applied to AI, the principle asks adults to take responsibility for the conditions they create. It does not supply an automatic verdict on a particular product.

A church can express that responsibility through clear purpose, careful selection and a genuine route to human help. It should not imply that a child has been heard by a responsible adult when a message has only been processed by a system. Nor should it blame a child for misunderstanding a relationship the institution presented ambiguously.

These are ethical recommendations for religious organizations. They are not a claim that every Christian tradition uses the same framework or that UNICEF endorses this theological interpretation. The source framework and the Christian application remain distinct.

Christian perspective: a decision record a small institution can maintain

Before adoption, record the need, intended age group, information involved and person responsible for the decision. Note which questions were answered by supplier documentation and which remain unresolved. State what would cause the institution to narrow, change or end the use.

After a bounded trial, ask whether the intended benefit occurred and whether anyone encountered an avoidable barrier. Keep the account proportionate and protect the privacy of participants. A small institution can do this without pretending to have conducted a large research study.

The central lesson is that care should be built into the arrangement around the tool. Telling children to exercise judgment cannot replace the judgment of adults who introduce the service. A credible approach joins opportunity, protection, listening and responsibility, with evidence about the actual use rather than reassurance based on a product category.

Mark 10:13–16 depicts Jesus welcoming children whom others were turning away. The scene gives children attention and value rather than treating them as an inconvenience. It does not supply a modern product checklist, but it challenges institutions to take their experience seriously.

Christian care requires more than telling children to be cautious. Adults choosing tools for them should understand the tool, provide alternatives and make it easy to ask for help without shame.

Use the guidance as a structured discussion with parents, educators and young people. Ask how a tool serves a specific need and how the organization will notice and respond if it causes difficulties.