Table of Contents
Kids and AI Drawings: Polish Is Fine, Don't Change My Idea
A CHI 2026 study on kids and AI drawings found children 8–13 welcomed AI adding detail but resisted AI changing their ideas. Where they draw the line and why.
Children already know where “help” ends and “taking over” begins. When researchers watched kids and AI drawings take shape together, in a study presented at the 2026 CHI conference, children aged 8 to 13 were happy to let the AI add details to their pictures. What they pushed back on was the AI changing what was in the picture. Add shading to my dragon: fine. Turn my dragon into a dinosaur: no.
That line, between refining an idea and replacing it, is exactly where authorship researchers and even the US Copyright Office locate ownership. Kids got there on their own. The job for parents is to notice it, name it and not talk them out of it.
Key Takeaways
- Fusco, Pitt, Newman, Yip and colleagues (CHI 2026) ran a participatory design study with children aged 8 to 13 co-creating with AI.
- Children “were open to technical AI refinements (e.g., adding details to their drawings) as scaffolds for developing drawing skills, but resisted conceptual transformations (e.g., changing objects) that altered their original ideas.”
- An earlier University of Washington study (CHI 2024) saw kids 7 to 13 sometimes abandon their own ideas when AI fell short, which researchers flagged as a concern.
- Adult research matches the kids’ instinct: people feel more ownership of AI-assisted text the more they influence it (Draxler et al., 2024).
- At home, ask “Is the AI helping your idea, or changing it?” and give kids the right to say no to the second.
What the CHI 2026 Study on Kids and AI Drawings Found
Child-AI co-creation means a child and a generative AI system make something together, like a story, a drawing or a comic, taking turns or sharing the work. It’s now common in kids’ apps and classroom tools. How children actually negotiate that collaboration has barely been studied.
Francesca Fusco, Caroline Pitt, Michele Newman, Jason Yip and Nikita Soni addressed that gap in “Towards Understanding Children’s Collaborative Interaction Patterns in Child-AI Co-creative Interfaces,” published in the Proceedings of the 2026 CHI Conference on Human Factors in Computing Systems on April 13, 2026. CHI is the leading peer-reviewed venue in human-computer interaction.
They ran a participatory design study, a method where children work alongside researchers as design partners rather than test subjects. The study looked at the roles children and AI took and the strategies children used to “align AI’s output with their intent.”
From the paper’s abstract, the key findings:
- Four collaboration profiles. Children didn’t all work with the AI the same way. The researchers identified four distinct child-AI collaboration patterns.
- A clear line on what AI may change. Kids accepted technical refinements, such as added details, and saw them as scaffolds for building their own drawing skills. They resisted conceptual changes, such as swapping objects, that altered their original idea.
- A design framework. The authors propose the Child-Centered Co-creative AI framework (“Kai”), grounded in how children naturally behave during co-creation.
A note on what we could verify: we read the published abstract and metadata. The full paper sits behind a security wall we couldn’t pass, so we’re not reporting sample size, session details or the names of the four profiles. Treat the specifics here as limited to what the abstract states.
Why Kids Draw the Line at Their Ideas
Here’s a way to think about the mechanism. Any creative work has two layers. The conceptual layer is the idea: what’s in the picture, what happens in the story, what it’s about. The execution layer is how well that idea is rendered: line quality, color, spelling, shading.
When an AI works on the execution layer, the idea is still the child’s. The child can look at the result and recognize their own intention, done better. When the AI works on the conceptual layer, the result reflects the model’s choices. Generative image models produce what is statistically likely given a prompt and their training data, so a conceptual edit tends to pull a child’s unusual idea toward something more typical. The kid looks at it and sees someone else’s picture.
Kids in the study seem to have felt that difference directly. And the research on adults suggests the feeling is well founded.
Ownership tracks influence
Fiona Draxler and colleagues (ACM Transactions on Computer-Human Interaction, 2024) found in two studies that people “do not consider themselves the owners and authors of AI-generated text,” and that “higher levels of participants’ influence on texts increased their sense of ownership.” In a 36-person study at the 2024 Creativity and Cognition conference, Barrett Anderson, Jash Shah and Max Kreminski found ChatGPT users generated more detailed ideas but “felt less responsible for the ideas they generated.”
The law draws the same line
The US Copyright Office’s January 2025 report on AI and copyrightability concluded that “neither the use of AI as an assistive tool nor the incorporation of AI-generated content into a larger copyrightable work affects the availability of copyright protection for the work as a whole,” while prompts alone don’t make someone the author of the output. Human authors keep rights in their expressive inputs that are perceptible in the output and in creative modifications. In other words: AI polishing your work leaves it yours; AI generating the expression doesn’t. We explain the rules for kids in who owns what AI makes.
When kids give up their ideas instead
The resistance in the CHI 2026 study isn’t guaranteed. In a 2024 University of Washington study of 12 children aged 7 to 13 using ChatGPT, DALL-E and Google’s Magenta, lead author Michele Newman described a “red flag”: some kids would “try an idea and quickly say, ‘It’s not working, so I’m just going to change the idea.’” The same kids, by the end of the study, were saying things like “I don’t think this really expresses what I’m saying.” Children can learn to defend their ideas, but they don’t always do it by default, especially when the tool can’t produce what they imagined.
A September 2026 scoping review in Frontiers in Psychology (Niu et al.) listed “loss of personal voice” and “over-scaffolding” among the risks reported across 22 studies of generative AI and children’s creativity. The kids in the CHI study were, in effect, guarding against both.
Help vs. Takeover: A Decision Table for AI in Kids’ Creative Work
| What the AI does | Layer | Whose idea is it after? | Reasonable at home? |
|---|---|---|---|
| Adds shading, texture or detail to a child’s drawing | Execution | Child’s | Usually yes, and kids in the CHI study saw it as a way to learn |
| Fixes spelling and punctuation in a story | Execution | Child’s | Yes |
| Suggests three possible endings; child picks or ignores | Mixed | Child’s, if they choose and rewrite | Yes, with the child deciding |
| Replaces an object in the drawing with a different one | Conceptual | Shifts toward the AI | Only if the child asks for it |
| Rewrites the plot “to make it better” | Conceptual | AI’s | Rarely; the story stops being theirs |
| Generates the whole image from a one-line prompt | Both | AI’s | Fine as play, but not “your drawing” |
How to Respect the Line Your Kid Already Draws
Ask the help-or-change question
When your child uses an AI drawing or story tool, ask: “Is it helping your idea, or changing it?” This gives kids words for what the CHI 2026 participants did instinctively. Most children answer quickly and accurately.
Give them veto power
If the AI changes something conceptual, your child gets to undo it, no debate. In a lot of kids’ apps, the AI’s version replaces the child’s with one tap. Show them how to go back, or keep a screenshot of the original so the idea isn’t lost.
Start on paper
Have your child sketch or write the core idea before any AI touches it. A paper sketch is a fixed record of their concept. If the AI drifts, they have something to compare to. This is the logic behind keeping kids as authors when AI illustrates, which we cover in kids write, AI illustrates.
Treat “the AI couldn’t do it” as a feature
When the tool can’t draw what your child imagined, that’s the moment the UW researchers saw kids give up their ideas. Flip it: “The AI can’t do your idea. That means it’s original. Let’s draw it ourselves.” Unusual ideas being hard for a model to render isn’t a bug in your kid.
What not to do
Don’t “improve” your child’s idea for them, with or without AI. Parents do this with the best intentions: a better title, a more logical plot. The research on ownership suggests every conceptual edit from someone else makes the work feel less like theirs. Offer execution help freely. Offer idea help only when asked.
What to Watch For Over the Next 3 Months
- Week 4: Your child can explain which parts of an AI-assisted project were their idea and which were the tool’s. They’ve used the undo or “go back” option at least once.
- Month 2 red flags: Your child routinely accepts whatever the AI produces. They describe projects as “what the AI made.” They drop ideas as soon as the tool can’t render them.
- Month 3 self-check: Ask your child to show you something they made with AI help and tell you the original idea. If the idea is clearly visible in the result and they’re proud of it, the line is holding.
Frequently Asked Questions
What did the CHI 2026 study on kids and AI drawings find?
Researchers ran a participatory design study with children aged 8 to 13 co-creating with AI. Kids accepted technical refinements, like the AI adding details to their drawings, and saw them as help for building skills. They resisted conceptual changes, like the AI swapping objects, that altered their original ideas. The authors identified four collaboration profiles and proposed a “Kai” design framework.
Is it OK for my child to use AI to improve their drawings?
Generally yes, when the AI works on how the idea looks rather than what the idea is. Children in the CHI 2026 study saw added details as a way to develop drawing skills. Keep the original, let your child decide which changes stay, and make sure the core idea remains theirs.
Why does my child get upset when the AI changes their picture?
Because it may no longer feel like theirs. Research on adults shows sense of ownership rises with personal influence over the work. When an AI swaps objects or rewrites ideas, the result reflects the model’s choices. Your child’s frustration is a healthy authorship instinct, not fussiness.
Should I worry if my child just accepts whatever the AI makes?
Watch for it. A 2024 University of Washington study saw some kids abandon their ideas when the AI couldn’t produce them, which the researchers flagged as a concern. Ask what their original idea was and whether the result matches. Kids in that study got better at defending their ideas with support from adults and peers.
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
- 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
- University of Washington News. (2024, May 29). “Q&A: How AI affects kids’ creativity” (interview with Michele Newman on the CHI 2024 study). https://www.washington.edu/news/2024/05/29/ai-kids-creativity-chatgpt/
- Draxler, F., Werner, A., Lehmann, F., Hoppe, M., Schmidt, A., Buschek, D., & Welsch, R. (2024). “The AI Ghostwriter Effect: When users do not perceive ownership of AI-generated text but self-declare as authors.” ACM Transactions on Computer-Human Interaction. https://doi.org/10.1145/3637875
- Anderson, B. R., Shah, J. H., & Kreminski, M. (2024). “Homogenization effects of large language models on human creative ideation.” Creativity and Cognition 2024. https://doi.org/10.1145/3635636.3656204
- U.S. Copyright Office. (2025, January 29). “Copyright and Artificial Intelligence, Part 2: Copyrightability.” https://www.copyright.gov/ai/Copyright-and-Artificial-Intelligence-Part-2-Copyrightability-Report.pdf
- 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. https://www.frontiersin.org/journals/psychology/articles/10.3389/fpsyg.2026.1880052/full
