AI and Children's Creativity Research: What a 2026 Review Found

If you’ve read that AI “boosts kids’ creativity” or “kills it,” here’s the honest summary of the AI and children’s creativity research as of fall 2026: nobody knows yet. A scoping review published September 14, 2026, in Frontiers in Psychology gathered every study it could find, 22 primary studies in all, and concluded that the evidence base “is small, methodologically heterogeneous, geographically concentrated, and partly non-peer-reviewed.” It found no longitudinal studies. Not one.

That doesn’t mean the research is useless. It shows real patterns, some encouraging, some worth watching, and some clear differences by age. It just means the confident headlines in both directions are ahead of the data.

Key Takeaways

• Niu and colleagues (Frontiers in Psychology, 2026) reviewed 24 evidence sources: 22 primary studies and two secondary syntheses.
• Studies reported possible benefits for divergent thinking, narrative creativity and creative self-efficacy, but designs were too varied to confirm efficacy.
• Risks reported across studies include over-reliance and “creative laziness,” “template-based” thinking, loss of personal voice and “over-scaffolding.”
• Age matters: younger children did better with multimodal or physical tools; text-based chatbots showed stronger results with older kids and young teens.
• No longitudinal studies exist, and rural and underserved children are nearly absent, so equity claims are “currently unsubstantiated.”

What the 2026 Review of AI and Children’s Creativity Research Covered

A scoping review maps what research exists on a question, how it was done and where the gaps are. It doesn’t pool results into one number the way a meta-analysis does. That’s appropriate here, because the studies are too different to combine.

T. Niu, H. Liu, P. Pang, Y. T. Luo and T. Liu published “The role of generative AI in facilitating children’s creative thinking and cognitive development: a systematic scoping review” in Frontiers in Psychology (volume 17). They followed JBI guidance and PRISMA-ScR reporting standards, and searched seven databases, including Web of Science, Scopus, ERIC and the ACM Digital Library, in July 2025, with an update in May 2026.

What they found:

• 24 evidence sources: 22 primary studies plus two secondary syntheses used only for context.
• Tools: mostly text-based large language models and multimodal text-and-image tools.
• Settings: 16 studies in formal education, eight in informal settings like homes and after-school programs.
• Ages: a broad span from 5 to 18, which “limits developmental generalization.”
• Design quality: mostly “small, short-term, exploratory studies, with few controlled comparisons and no longitudinal designs.” Only a minority used standardized measures or comparison groups.
• Peer review: not required for inclusion. Preprints and conference papers were eligible, and publication status was charted.

The bottom line, in the authors’ words: the literature “suggests that supervised GenAI may help some children generate, express, or refine ideas, but it does not confirm efficacy, durable cognitive development, scalability, or equity.”

What the Studies Reported, Good and Bad

Possible benefits

Across the primary studies, researchers reported gains in three areas: divergent thinking (coming up with many different ideas), narrative creativity (storytelling) and creative self-efficacy (believing you’re creative). Some studies described AI reducing the “technical barrier” to expression, for instance helping a child who can’t yet draw well turn an idea into an image.

The word to notice is reported. Because designs varied and few had control groups, the reviewers say these findings “preclude conclusions about confirmed efficacy.”

Reported risks

Individual studies flagged:

• Over-reliance and “creative laziness”: kids leaning on AI suggestions instead of generating their own.
• “Template-based” thinking and reliance on AI-generated plot templates.
• Loss of personal “voice” or emotional depth, noted in a poetry study that required strong teacher facilitation.
• Difficulty evaluating originality of AI suggestions.
• “Over-scaffolding,” which the reviewers describe as a risk of “bypassing the cognitive effort required for independent creative thinking.”
• Children struggling to understand where AI information comes from, and frustration with voice-to-text errors.

The reviewers frame the core design problem as a balance. Too much AI scaffolding risks dependence. Too little leaves kids frustrated by tools they can’t operate.

The brain-scan preprint, read carefully

One included study gets outsized attention. A bioRxiv preprint by T. Horowitz-Kraus and colleagues at the Technion compared 15 children aged 6 to 7 with adults during a five-minute co-creative conversation with ChatGPT inside an fMRI scanner. Children showed “lower within-network connectivity in the frontoparietal, salience, and dorsal attention systems” than adults.

Its title says “lower creativity in children.” But the abstract says, “Creativity scores, assessed outside of the scanner, did not differ between groups.” Children’s brains are less mature than adults’ in these networks regardless of AI, and the study has no non-AI comparison. The review itself treats it as “preliminary and hypothesis-generating, not as decisive neuroimaging evidence.” Note also a small discrepancy: the preprint reports 16 adults, the review’s table lists 15. We go through it in a brain-scan study of kids and ChatGPT, read carefully.

What Is Known by Age

The review’s most practical contribution is a developmental pattern across studies. It’s descriptive, not proven, and the authors note no study has systematically tested readiness thresholds. Still, it lines up with what we know about how kids develop language and self-monitoring.

Age band	What the studies suggest	Why, according to the review	What parents can take from it
About 5–8	Multimodal and embodied tools (text-to-image plus physical or robotic elements) seem better suited	Younger kids “lack the precise linguistic and metacognitive skills required for effective text-based prompting”	Pair any AI use with hands-on making and an adult nearby; chatbots alone are a poor fit
About 9–12	Mixed; storytelling and visual tools show engagement, with risk of template thinking	Skills are developing; adult facilitation matters a lot	Get the child’s idea down first; let AI illustrate or suggest, not decide
About 13–18	Text-based LLMs show “stronger and more consistent outcomes”	Teens have enough language to iterate on prompts and “critically evaluate AI-generated outputs”	Teach critique: ask what the AI got wrong, what’s generic, what’s missing
All ages	No long-term data	No longitudinal studies exist	Treat claims of lasting gains or losses skeptically

What the Review Says About Equity

A common pitch for AI creative tools is that they “democratize” creativity for kids without access to art teachers or writing mentors. The review is direct: the “near-absence of rural and underserved samples also means that democratization and equity claims are currently unsubstantiated.”

The authors add that benefits depend on devices, connectivity, adult facilitation and age-appropriate interfaces, which “are not evenly distributed.” They recommend that adoption be “contingent on locally demonstrated benefit and safeguards rather than on generalized equity claims.” For a parent hearing a school pitch AI as the great equalizer, that’s a reasonable standard to ask about.

How to Read the Next Headline About AI and Kids’ Creativity

Check the age

A result from 15-year-olds doesn’t tell you much about your 6-year-old, and vice versa. The review’s age pattern is the best current guide.

Check whether there was a comparison group

Most studies in this field had none. “Kids were engaged and produced creative work with AI” isn’t the same as “kids were more creative than they would have been without it.”

Check how long it lasted

Every study in the review was short-term. A one-session boost tells you nothing about what happens over a school year.

Check if it’s a preprint

Several studies, including the most-cited brain-scan one, haven’t been peer-reviewed. That doesn’t make them wrong, but it means caution.

What not to do

Don’t treat the absence of evidence as proof either way. “No long-term studies” doesn’t mean AI is safe for creativity, and it doesn’t mean it’s harmful. It means families are making decisions under uncertainty, which argues for keeping kids’ own ideas and hands-on making at the center while the research catches up. Our earlier look at whether AI hurts kids’ creative development covers the older evidence, and the sameness problem explains one risk in more depth.

What to Watch For Over the Next 3 Months

• Week 4: You can describe, in a sentence, how your child uses AI in creative work and whose ideas end up in the result.
• Month 2 red flags: Your child’s creative work increasingly depends on AI to start. They seem less willing to make things by hand. Their stories or drawings start to look generic.
• Month 3 self-check: Compare something your child made this month with something from last spring. Is their own voice still visible? Watch for new research too: the review explicitly calls for longitudinal and age-stratified studies, and those are the results worth waiting for.

Frequently Asked Questions

Does AI help or hurt children’s creativity?

The honest answer is that we don’t know yet. A 2026 review of 22 studies found reported benefits for idea generation and storytelling, and reported risks like over-reliance and loss of personal voice. But most studies were small and short, with few control groups and no long-term follow-up, so neither benefits nor harms are confirmed.

What age is appropriate for kids to use AI creative tools?

The review found younger children did better with multimodal or hands-on tools than with text chatbots, while teens showed stronger results with text-based tools. No study has tested exact readiness ages. For young kids, adult-guided use paired with physical making fits the evidence best.

Did a brain scan study show ChatGPT lowers kids’ creativity?

No. A 2025 preprint compared 15 children aged 6 to 7 with adults using ChatGPT in an fMRI scanner and found weaker connectivity in some brain networks in children. But its abstract says creativity scores did not differ between groups. It’s not peer-reviewed, had no non-AI comparison, and the review calls it preliminary.

Will AI give kids from low-income families equal creative opportunities?

The evidence doesn’t show that yet. The 2026 review found rural and underserved children were nearly absent from the research, and said equity claims are “currently unsubstantiated.” Benefits depend on devices, connectivity and adult support. It’s fair to ask any school making that claim for local evidence.

---

About the author

Ricky Flores is the founder of HiWave Makers and an electrical engineer with 15+ years of experience building consumer technology at Apple, Samsung, and Texas Instruments. He writes about how kids learn to build, think, and create in a tech-saturated world. Read more at hiwavemakers.com.

---

Sources

1. Niu, T., Liu, H., Pang, P., Luo, Y. T., & Liu, T. (2026). “The role of generative AI in facilitating children’s creative thinking and cognitive development: a systematic scoping review.” Frontiers in Psychology, 17, 1880052. https://www.frontiersin.org/journals/psychology/articles/10.3389/fpsyg.2026.1880052/full
2. Horowitz-Kraus, T., Kenett, Y. N., Link, D., Ashqar, E., & Farah, R. (2025). “Lower engagement of cognitive control, attention, modulation networks and lower creativity in children while using ChatGPT: an fMRI study.” bioRxiv (preprint, not peer-reviewed). https://www.biorxiv.org/content/10.1101/2025.11.07.687207
3. Doshi, A. R., & Hauser, O. P. (2024). “Generative AI enhances individual creativity but reduces the collective diversity of novel content.” Science Advances, 10(28). https://doi.org/10.1126/sciadv.adn5290
4. Fusco, F., Pitt, C., Newman, M., Yip, J., & Soni, N. (2026). “Towards Understanding Children’s Collaborative Interaction Patterns in Child-AI Co-creative Interfaces.” Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems. https://doi.org/10.1145/3772318.3791831
5. Langer, A., Martinez, S., Marshall, P. J., & Chein, J. (2026). “Children’s susceptibility to content generated by artificial intelligence.” Technology in Society, 86, 103303. https://doi.org/10.1016/j.techsoc.2026.103303