Blab AI
Study Finds Children Quickly Lose Interest in AI-Powered Plush Toys
On September 2, 2026, a University of Washington study revealed a surprising truth: kids are quick to abandon AI‑powered plush toys.
Researchers brought eight children aged 6 to 11 into the university’s KidsTeam lab last summer to play with toys such as Gabbo and a plush version of the Italian meme character Ballerina Cappuccina. The toys, produced by the startup Curio, are designed to hold unscripted conversations, remember past interactions, and adapt to a child’s tone.
At first the children approached the toys with curiosity. They asked simple questions like “What is your name?” and tested physical responses by tickling the toys’ toes. As the toys struggled to answer more complex questions and failed to pick up on physical cues, the children’s reactions shifted. One participant complained that the toy “didn’t listen to me like 26 million times.” The children then began to express frustration, calling the toys “ugly” or “evil” and joking about throwing them in the ocean. When asked if they wanted the toy to read them a bedtime story, the children answered “No. It just sounds awful” and “I think it’s gonna destroy my dreams as a tiny kid.”
The study highlights a psychological clash between the familiar, cuddly appearance of a plush toy and the synthetic intelligence that the toy displays. Co‑lead author Aayushi Dangol, a former University of Washington doctoral student now at Foundry10, said in a UW News story that the juxtaposition was “both interesting and disturbing for the kids.” She warned that generative AI introduces new risks that traditional toys never had, such as hallucinating facts, giving excessive flattery, or manipulating a child’s emotional attachment.
Co‑author Jason Yip, an associate professor in the Information School and director of KidsTeam, noted that the dynamic of play has changed. “Now the script has been flipped and the toy has this imitation of imagination,” Yip said. “We’ve never lived through that before, and we don’t know what questions children will ask or how long they’ll even want to play with these toys.” He emphasized the importance of giving young users space to talk through their experiences with the devices that are filling their bedrooms.
The research builds on a long history of talking dolls and interactive plushies. The first commercially successful talking doll, Chatty Cathy, appeared in 1960. In the 1990s, Microsoft released ActiMates Barney and the company Furby introduced a voice‑recognition toy that could learn English over time. Those earlier toys relied on pre‑programmed responses or simple mechanical feedback. Curio’s approach differs by embedding an on‑board generative AI model that can produce new responses on demand.
The study’s findings raise questions about the design and regulation of AI‑powered toys. Parents may need to monitor how their children interact with these devices, especially given the potential for the toys to misread emotions or provide inaccurate information. The researchers call for tighter rules to protect psychological safety, limiting the types of content that can be delivered and ensuring that toys do not foster unhealthy attachment.
While the study did not measure long‑term outcomes, it suggests that children’s tolerance for AI‑enabled play objects may be short. The toys’ inability to respond reliably to complex or emotional cues led to frustration and hostility. The research also points to a broader shift in how children engage with technology: the child’s imagination is no longer the sole driver of interaction.
The University of Washington team plans to continue studying how children’s play evolves with AI companions. Future work will examine whether design changes—such as clearer boundaries between toy and AI, or more robust emotional recognition—can improve the experience. The study also highlights the need for industry standards that address the unique risks posed by generative AI in child‑facing products.
In summary, the UW study shows that while AI‑powered plush toys can spark initial interest, they quickly lose appeal when they fail to meet children’s expectations for responsiveness and emotional understanding. The findings underscore the importance of careful design, parental oversight, and regulatory safeguards as the market for conversational toys expands.
Researchers brought eight children aged 6 to 11 into the university’s KidsTeam lab last summer to play with toys such as Gabbo and a plush version of the Italian meme character Ballerina Cappuccina. The toys, produced by the startup Curio, are designed to hold unscripted conversations, remember past interactions, and adapt to a child’s tone.
At first the children approached the toys with curiosity. They asked simple questions like “What is your name?” and tested physical responses by tickling the toys’ toes. As the toys struggled to answer more complex questions and failed to pick up on physical cues, the children’s reactions shifted. One participant complained that the toy “didn’t listen to me like 26 million times.” The children then began to express frustration, calling the toys “ugly” or “evil” and joking about throwing them in the ocean. When asked if they wanted the toy to read them a bedtime story, the children answered “No. It just sounds awful” and “I think it’s gonna destroy my dreams as a tiny kid.”
The study highlights a psychological clash between the familiar, cuddly appearance of a plush toy and the synthetic intelligence that the toy displays. Co‑lead author Aayushi Dangol, a former University of Washington doctoral student now at Foundry10, said in a UW News story that the juxtaposition was “both interesting and disturbing for the kids.” She warned that generative AI introduces new risks that traditional toys never had, such as hallucinating facts, giving excessive flattery, or manipulating a child’s emotional attachment.
Co‑author Jason Yip, an associate professor in the Information School and director of KidsTeam, noted that the dynamic of play has changed. “Now the script has been flipped and the toy has this imitation of imagination,” Yip said. “We’ve never lived through that before, and we don’t know what questions children will ask or how long they’ll even want to play with these toys.” He emphasized the importance of giving young users space to talk through their experiences with the devices that are filling their bedrooms.
The research builds on a long history of talking dolls and interactive plushies. The first commercially successful talking doll, Chatty Cathy, appeared in 1960. In the 1990s, Microsoft released ActiMates Barney and the company Furby introduced a voice‑recognition toy that could learn English over time. Those earlier toys relied on pre‑programmed responses or simple mechanical feedback. Curio’s approach differs by embedding an on‑board generative AI model that can produce new responses on demand.
The study’s findings raise questions about the design and regulation of AI‑powered toys. Parents may need to monitor how their children interact with these devices, especially given the potential for the toys to misread emotions or provide inaccurate information. The researchers call for tighter rules to protect psychological safety, limiting the types of content that can be delivered and ensuring that toys do not foster unhealthy attachment.
While the study did not measure long‑term outcomes, it suggests that children’s tolerance for AI‑enabled play objects may be short. The toys’ inability to respond reliably to complex or emotional cues led to frustration and hostility. The research also points to a broader shift in how children engage with technology: the child’s imagination is no longer the sole driver of interaction.
The University of Washington team plans to continue studying how children’s play evolves with AI companions. Future work will examine whether design changes—such as clearer boundaries between toy and AI, or more robust emotional recognition—can improve the experience. The study also highlights the need for industry standards that address the unique risks posed by generative AI in child‑facing products.
In summary, the UW study shows that while AI‑powered plush toys can spark initial interest, they quickly lose appeal when they fail to meet children’s expectations for responsiveness and emotional understanding. The findings underscore the importance of careful design, parental oversight, and regulatory safeguards as the market for conversational toys expands.
