Researchers at the University of Cincinnati invited 41 children, aged 5 to 10, to play three games with a toy robot called Miko. Miko is commercially available and runs on generative AI. It talks, tells stories, answers questions and has an animated face.

Afterwards, the researchers asked the children what they thought.

Two findings stand out.

They believed it. Per UC's release, the children "tended to believe what the AI robot told them was true." They were enthusiastic, including those who had never seen a robot like it before.

At five to ten, that isn't surprising. Young children extend trust to things that talk to them kindly and confidently. But it's worth saying plainly, because generative AI is confident about everything, including the things it gets wrong.

Some of them called it a friend. The researchers asked whether Miko was "a toy, a friend or something else." Nancy Jennings described what stood out:

Read that twice. A child who is struggling with other children finds something that is always patient, always kind and never rejects them. They love it. And they tend to believe what it says.

That's not a child being fooled. It's a child finding something gentle in a hard week, from a machine with no stake in whether they're okay.

The study's own headline is "Unlike many parents, children are fearless around AI," and Kelly Merrill's biggest takeaway was simply: "these children are extremely excited about robots and AI... AI can be a scary thing for parents."

The researchers credit real benefits. Jennings notes these toys can help children "practice conversation and improve vocabulary and listening skills," and that they "might appeal to some children who have limited social outlets outside of school."

They also watched Miko's safety features work. During its Story Maker game:

The broader concerns in the release are about AI toys as a category, not Miko. It cites a Common Sense Media risk assessment this year that found AI toys "regularly generated age-inappropriate content," and privacy concerns raised by the Public Interest Research Group. Those are worth knowing before you buy anything. They aren't findings about this robot.

This is small and early. 41 children, qualitative interviews, presented at the Comm Horizons conference at UC Davis and described in a university press release. There's no published paper yet and no figures for how many children believed what, or how "believing" was measured.

That doesn't make it useless. Qualitative work like this is how researchers find out what to measure next. But it's an observation, not a proven result, and we'd be doing you a disservice to present it any other way.

One more detail that deflates a common headline: half the children "regularly used AI agents at home," and Merrill says "the most common response I got was that they used AI for music: 'Hey, Siri, play this song' or 'Hey, Alexa.'" So these are mostly voice assistants, not five-year-olds with chatbots.

AI literacy, early. Jennings: "we will have to educate children at a young age to use critical thinking about the tools and talk through some of their experiences with their parents."

UC's own summary defines AI literacy as knowing "what artificial intelligence can and can't do," being sceptical of its answers, and understanding that "generative AI can be manipulative because it prizes agreeableness over facts." That last phrase is from the university's FAQ, not the researchers, but it's the clearest one-line explanation of the problem we've seen.

Not a replacement. Merrill: "I think it's important to understand that these AI tools are not replacements — they're complements. We shouldn't be looking to replace people."

Sit in, sometimes. Not to supervise every minute, but often enough to hear what the toy says and how your child responds.

Ask one question, often: "How would it know that?" It's the simplest AI literacy lesson there is. It teaches that answers come from somewhere, and that "it said so" isn't a source.

Catch it being wrong together. Ask the toy something you know the answer to, and when it slips (it will, eventually), make it a small discovery rather than a warning. "Huh, it got that wrong. Robots do that." A child who has seen it make one mistake believes it differently afterwards.

Treat "it's my friend" as information about school. If your child says the robot is a friend who doesn't judge them, that's not mainly a fact about the robot. It may be telling you something about the playground. Follow up on that.

Keep it in shared spaces. A robot in the living room is part of family life. A robot alone in a bedroom at night is a different relationship.

Check the settings before you buy. Privacy, data retention, parental controls and what happens to recordings. Merrill's joke in the release, as Siri interrupted him from his desk, is worth remembering: "Robots are always listening."

The children in this study weren't confused or reckless. They were excited, trusting and, in some cases, a bit lonely. That's what young children are like.

The job isn't to take the robot away. It's to make sure the robot isn't the only one listening.

Source: University of Cincinnati, "Unlike many parents, children are fearless around AI," September 2026. https://www.uc.edu/news/articles/2026/09/uc-study-children-ai-toys-ai-literacy.html

A university press release about findings presented at the Comm Horizons conference. 41 children, qualitative; no peer-reviewed paper or quantitative results have been published. Quotations are verbatim. Lines from the page's FAQ are attributed to the University of Cincinnati rather than the researchers. The concerns cited from Common Sense Media and PIRG concern AI toys generally, not Miko. The guidance for parents is ours.

Disclosure: drafted with Claude, made by Anthropic.

Want practical AI guidance for parents and educators every week? Subscribe: https://www.aibyage.com/?modal=signup&utm_source=beehiiv&utm_medium=newsletter