Research on generative artificial intelligence and learning has focused overwhelmingly on adults and university students, according to a 2026 mini review published in Frontiers in Psychology.
Only eight of 46 studies examined anyone under 18. Thirty-five of the studies that collected primary data sampled university students. The review was led by Yaoyao Fan and colleagues and used self-determination theory to consider how children might be supported as they use generative AI.
A major evidence gap
The review was a short, theory-guided narrative review rather than a full systematic search. That means its 46 studies should not be treated as a complete count of all research in the field.
Even within that selected group, evidence about children was limited. Twenty-six studies were cross-sectional surveys, while four followed participants over time and two were experiments. Many of the underlying studies relied on self-report questionnaires, which may not fully capture how learners use AI in practice.
None of the reviewed studies measured guidance at home and at school in the same participants. Four studies examined a family factor, and 15 examined a school factor, but no study measured both settings together.
As a result, claims about children in the review’s framework are extrapolations from older learners. The proposed model for coordinating support between families and schools is a plan for future testing, not an observed result from the reviewed research.
Why the type of guidance may matter
Self-determination theory identifies autonomy, competence and relatedness as basic psychological needs. Applied to AI use, the review describes autonomy support as guidance that gives children meaningful choices, considers their perspective and explains the reasons behind rules.
Autonomy support is not presented as a lack of boundaries. The review pairs it with structure: clear expectations, help developing skills and firm limits.
One study of 1,036 high school students found that students who viewed their teachers as both autonomy-supportive and clear about expectations had the strongest motivation and self-regulated learning outcomes. Among German secondary students in classrooms using ChatGPT, perceived teacher support was the strongest predictor of a healthy motivational and emotional profile.
Other findings point in a similar direction, while remaining limited by the ages and designs of the studies. AI self-study rooms were linked to gains in motivation, self-regulation, enjoyment and engagement among primary-school learners.
Use may be more important than frequency
The reviewed evidence suggests that how learners use AI may matter more than how often they use it. However, the relevant studies involved university students or working adults rather than a broad sample of children.
In a three-wave study of 589 university students and early-career knowledge workers, dependent AI use was linked to lower intrinsic motivation and poorer perceived thinking outcomes. Autonomous use showed the opposite pattern. Both types of use produced the same immediate benefit in that study.
Another study involving 72,615 Chinese university students linked AI use with greater cognitive and emotional engagement, but also with lower academic motivation and less active learning. These findings describe associations in university populations; they do not establish that AI caused the outcomes or that the same pattern applies to children.
What family research adds
The review included four family studies. They were qualitative, cross-sectional or mixed-method studies, and three surveyed or interviewed parents rather than children.
Interviews with 16 parents of young children across nine Chinese cities suggested that parents’ beliefs influenced AI’s role at home, including how they managed screen time and content. In Türkiye, a survey found that 28.4% of parents of children with speech and language disorders had used ChatGPT for health information; use varied by gender and education.
Families managing a child’s chronic illness preferred a condition-specific chatbot to general-purpose ChatGPT. This finding concerns family preferences and does not establish that either type of tool is clinically appropriate.
Among Chinese middle-school students, independent AI use and family discussion were associated with stronger endorsement of AI ethics principles in the reviewed study context. The evidence described here does not establish causation.
What families and schools can take from the review
The review does not provide a tested formula for managing children’s AI use. Its practical value is in identifying questions that current research has not answered.
- Set expectations clearly: Children may need understandable boundaries around when and why AI can be used.
- Explain the reasons: Guidance that includes a child’s perspective and the rationale for rules fits the review’s description of autonomy support.
- Build skills: Structure includes help developing the ability to use AI thoughtfully, rather than relying on restrictions alone.
- Connect home and school: Researchers have not yet tested whether coordinated guidance produces additive, synergistic or compensatory effects.
- Look beyond self-report: Future studies need to examine behavior as well as what learners say about their AI use.
The central message is one of caution. Guidance for children’s generative AI use is developing faster than the evidence base. More research involving the same children across home and school settings will be needed before firm conclusions can be drawn about which approaches work best.
AI tools were used to assist with the preparation of this article.
Leave a Reply