Kids ChatGPT Brain Study: A Brain-Scan Preprint, Read Carefully

A kids ChatGPT brain study posted in November 2025 has a title built to travel: “Lower engagement of cognitive control, attention, modulation networks and lower creativity in children while using ChatGPT: an fMRI study.” Read the abstract and you find this sentence: “Creativity scores, assessed outside of the scanner, did not differ between groups.”

So did kids show lower creativity or not? According to the study’s own summary, not on the creativity test. And there’s a bigger issue the title hides: the study compared children with adults, both using ChatGPT. There was no group of kids who didn’t use ChatGPT. It can’t tell you what ChatGPT does to a child’s brain. It can tell you that 6-year-old brains work differently from adult brains, which we already knew.

Key Takeaways

• The preprint (Horowitz-Kraus et al., bioRxiv, November 2025) scanned 15 children aged 6 to 7 and 16 adults during a five-minute co-creative chat with ChatGPT.
• Children showed lower connectivity in control and attention networks than adults. The abstract says creativity scores “did not differ between groups,” despite “lower creativity” in the title.
• There was no condition without ChatGPT, so the study can’t show ChatGPT caused any difference.
• It’s a preprint, not peer-reviewed, with small samples, a known weakness in brain-imaging research.
• The useful lesson for families is how to read a study: title, abstract, comparison group, sample size, peer review.

What the kids ChatGPT brain study actually did

The study is a preprint on bioRxiv by Tzipi Horowitz-Kraus, Yoed Kenett, D. Link, E. Ashqar and Rola Farah, posted November 10, 2025 (version 2 on November 12). As of this writing, bioRxiv lists no published journal version.

Here’s the design, from the abstract:

• Participants: 15 children aged 6 to 7, and 16 adults.
• Task: “a naturalistic 5-minute co-creative dialogue with ChatGPT using the same prompt,” held inside an fMRI scanner.
• Measure: functional connectivity within and between seven large brain networks, including the frontoparietal (control), salience, dorsal attention and default mode networks.
• Creativity: measured separately, “outside of the scanner.”

And the results:

• Adults showed “stronger within-network connectivity in control/attention systems,” with the clearest effect in the frontoparietal (FP) network, plus more integration across networks.
• “Creativity scores, assessed outside of the scanner, did not differ between groups.”
• In adults, but not children, higher creativity went with stronger connectivity in control networks.

The authors’ own interpretation is modest: “The combination of similar creativity scores with divergent neural profiles suggests that developmental differences in the implementation of creative cognition may be more pronounced than baseline capacity.” In plain terms: kids and adults scored similarly on creativity but used their brains differently during the AI chat.

Why the title and the findings don’t match

The creativity claim. The title says “lower creativity in children while using ChatGPT.” The abstract reports no difference in creativity scores. The likeliest reading is that “lower creativity” refers to the brain-network patterns the authors associate with creative cognition, not to a measured creativity deficit. Either way, a reader who sees only the title, which is how most people see research on social media, gets a claim the data summary doesn’t support.

The missing comparison. To learn what ChatGPT does to a child’s brain, you’d compare children using ChatGPT with children doing a similar task without it, or the same children both ways. This study compared children with adults. Control networks are still developing in childhood: a 2007 resting-state study in PNAS (Fair et al.) found that the adult frontoparietal and cingulo-opercular control networks form through development, as long-range connections strengthen. So finding that 6-year-olds show weaker control-network connectivity than adults is expected for any demanding task. The study doesn’t separate “child brain” from “child brain on ChatGPT.”

Small samples. Fifteen and sixteen people is small for fMRI. In a well-known 2013 analysis in Nature Reviews Neuroscience, Katherine Button and colleagues showed “the average statistical power of studies in the neurosciences is very low,” with consequences including “overestimates of effect size and low reproducibility of results.” Small studies can find real things. They also produce false alarms and exaggerated effects more often.

Not peer-reviewed. Preprints are valuable; they let science move fast. But no independent reviewers have checked this one yet. Peer review doesn’t guarantee a study is right, but it often catches exactly the kind of title-versus-findings mismatch seen here.

None of this means the researchers did bad work or that the question is unimportant. Studying young children’s brains during AI use is hard and worth doing. It means the headline claim outruns the evidence.

Why headlines get this wrong so often

Exaggeration usually starts upstream of journalists. Petroc Sumner and colleagues studied 462 press releases from 20 leading UK universities and the news stories based on them (BMJ, 2014). They found 33% of press releases contained exaggerated causal claims and 40% contained exaggerated advice. When the press release exaggerated, news stories followed: 81% of news stories made similar exaggerated causal claims, compared with 18% when the release didn’t.

A study title works like a mini press release. When it claims more than the abstract, headlines inherit the claim. That’s why the same skill that protects you from overhyped health news protects you from overhyped AI-and-brain news.

This isn’t the only AI brain study with a gap between attention and evidence. The MIT Media Lab’s “Your Brain on ChatGPT” preprint (Kosmyna et al., 2025), with adults writing essays, drew wide coverage and then a formal comment from other researchers raising concerns about “the limited sample size,” “the reproducibility of the analyses” and “methodological issues related to the EEG analysis.” We covered that one in what the MIT study found about ownership.

How to Teach Your Kid to Read a Science Headline

This study is a gift for teaching, because the gap between title and abstract is visible to anyone who reads both.

Ages 5–8: “how do you know?”

Make it a game at dinner. Someone says a claim (“carrots help you see in the dark”), and everyone asks “How do you know?” Then talk about who would need to be tested to find out. Young kids can grasp the core idea: to know if something changes people, you have to compare people who did it with people who didn’t.

Ages 9–12: the fair test

Describe this study simply: “Scientists put 15 kids and 16 grown-ups in a brain scanner while they talked to ChatGPT. Kids’ brains looked less connected than grown-ups’. Did ChatGPT cause that?” Let your kid work out that kids and adults differ anyway, and that a fair test needs kids without ChatGPT. Then have them design the fair test on paper: who’s in each group, what they do, what you measure.

Ages 13+: the five-question check

Give your teen this study’s title and abstract (both are free on bioRxiv) and have them answer five questions:

1. What does the title claim?
2. What does the abstract actually report?
3. Who was compared with whom?
4. How many people were in each group?
5. Has it been peer-reviewed?

Then have them write a more accurate headline. A good one might be: “Kids’ and adults’ brains work differently while chatting with ChatGPT, small study finds.” Next time a viral AI-brain headline shows up, they run the same five questions.

The question to ask: “If this study were true, what would it actually let us say about ChatGPT, and what wouldn’t it?”

The five-question check, applied to this study

Question	What to look for	This study
What does the title claim?	Strong causal words, “lower,” “damages,” “rewires"	"Lower engagement… and lower creativity in children while using ChatGPT”
What does the abstract report?	Whether the main result matches the title	Lower connectivity in kids vs adults; creativity scores “did not differ”
Who was compared with whom?	A group without the thing being tested	Children vs adults, both using ChatGPT; no no-ChatGPT group
How many people?	Larger is more reliable; small brain studies overestimate effects (Button et al., 2013)	15 children, 16 adults
Peer-reviewed?	Journal publication vs preprint	Preprint on bioRxiv; no published version listed
Verdict		Interesting early data on developmental differences; not evidence ChatGPT harms kids’ brains or creativity

What parents can do with studies like this

Read past the headline, every time

When a study about kids and AI goes viral, find the abstract. It’s usually free. Check whether the main result matches the claim. It takes two minutes and you’ll be the calmest parent in the group chat.

Don’t swing to “it’s fine,” either

A flawed study doesn’t prove the opposite of its claim. We still don’t know much about what AI use does to young children’s developing brains. Reasonable caution still applies: young kids using AI should do so with an adult, and it shouldn’t replace play, talk and hands-on activities. For broader context, see our explainer on AI assistance and kids’ brain development.

Watch for comparison groups in any claim

“Kids who use X show Y” means little without “compared with kids who don’t.” Train yourself to look for the comparison. Most misleading studies, and most misleading headlines, are missing one.

Share the lesson, not the panic

If someone forwards the headline, reply with the abstract sentence: “Creativity scores… did not differ between groups.” You’ll do more for other parents than another alarmed share.

What not to do

Don’t use this study to scare your kid off technology. Kids notice when adults overstate things, and it costs you credibility for the warnings that matter. Use it to teach them how to think instead.

What to Watch For Over the Next 3 Months

• Week 4: You’ve read the abstract of at least one viral study before reacting, and your teen has tried the five-question check once.
• Month 2 red flags: Your kid repeating alarming AI-brain claims as fact, or dismissing all research as fake. Both are signs the middle ground, “here’s what this one study can and can’t show,” needs more practice.
• Month 3 self-check: Can your teen explain why a study needs a comparison group? Can they find an abstract on their own? Watch whether this preprint gets a peer-reviewed version, and whether the title changes.

Frequently Asked Questions

Does using ChatGPT lower kids’ creativity?

This study doesn’t show that. Its title says “lower creativity,” but its abstract says creativity scores measured outside the scanner did not differ between the 6- to 7-year-olds and adults. And because there was no group without ChatGPT, it couldn’t test whether ChatGPT affects creativity at all.

What did the fMRI study of kids using ChatGPT find?

Fifteen children aged 6 to 7 showed lower connectivity in brain networks for control and attention than 16 adults during a five-minute chat with ChatGPT. That’s consistent with normal brain development, since those networks mature with age. The study is a preprint and hasn’t been peer-reviewed.

What is a preprint?

A preprint is a research paper shared publicly before peer review, often on sites like bioRxiv or arXiv. It lets findings circulate quickly, but independent experts haven’t checked the methods or conclusions yet. Treat preprint findings as early and tentative, and look for whether a peer-reviewed version appears later.

How can I tell if a study about kids and screens is trustworthy?

Ask five questions: what the title claims, what the abstract reports, who was compared with whom, how many people took part, and whether it’s peer-reviewed. Watch especially for missing comparison groups and small samples, and for headlines that turn associations into causes.

Should young kids use ChatGPT at all?

Many AI services set minimum ages of 13 or older, and UNESCO’s 2023 guidance suggested 13 as a minimum for classroom use. This study doesn’t show harm, but it doesn’t show safety either. For young children, any AI use should be with an adult and shouldn’t replace play and conversation.

---

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. 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
2. Button, K. S., Ioannidis, J. P. A., Mokrysz, C., Nosek, B. A., Flint, J., Robinson, E. S. J., & Munafò, M. R. (2013). “Power failure: why small sample size undermines the reliability of neuroscience.” Nature Reviews Neuroscience, 14(5), 365-376. https://doi.org/10.1038/nrn3475
3. Sumner, P., Vivian-Griffiths, S., Boivin, J., Williams, A., Venetis, C. A., Davies, A., Ogden, J., Whelan, L., et al. (2014). “The association between exaggeration in health related science news and academic press releases: retrospective observational study.” BMJ, 349, g7015. https://doi.org/10.1136/bmj.g7015
4. Kosmyna, N., et al. (2025). “Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task.” arXiv preprint, not peer-reviewed. https://arxiv.org/abs/2506.08872
5. Stankovic, M., Hirche, E., Kollatzsch, S., Doetsch, J. N., et al. (2025). “Comment on: Your Brain on ChatGPT.” arXiv. https://arxiv.org/abs/2601.00856
6. Fair, D. A., Dosenbach, N. U. F., Church, J. A., Cohen, A. L., Brahmbhatt, S., Miezin, F. M., Barch, D. M., Raichle, M. E., et al. (2007). “Development of distinct control networks through segregation and integration.” Proceedings of the National Academy of Sciences, 104(33), 13507-13512. https://doi.org/10.1073/pnas.0705843104
7. UNESCO. (2023, September 7). “UNESCO: Governments must quickly regulate Generative AI in schools.” https://www.unesco.org/en/articles/unesco-governments-must-quickly-regulate-generative-ai-schools