Harvard researchers didn't just study an AI toy this month — they built one. The Child-Centered AI Lab at the Harvard Graduate School of Education began lab testing its own AI-enabled toy against traditional, non-AI toys in early August 2026, marking one of the first controlled studies of how these devices actually change the way children play.
What is Harvard's Child-Centered AI Lab actually testing?
The lab, led by assistant professor Ying Xu, is comparing how children play with its purpose-built AI toy versus ordinary toys in a lab setting. Researchers are tracking play behavior, language use, social interaction, trust, and attachment. A second phase will move the study into participating families' homes for longer-term observation, though the lab hasn't yet said when that phase begins.
Xu's team also published a companion literature review, “Artificial Intelligence in Toys: Implications for Child Development and Play,” which surveys what's already known about how interactive toys shape kids' behavior — and it's the backdrop for the new study's questions.
What does earlier research say about AI toys and play?
The review Xu's team compiled draws on a handful of striking earlier findings. Scripted interactive dolls narrowed the range of what kids talked about. Talking action figures narrowed play themes compared to silent versions. And when a robot is simply given a human name, children trust it more.
| Toy or study | Documented effect |
|---|---|
| Hello Barbie | Scripted dialogue narrowed children's conversation range |
| Talking Rescue Heroes figures | Narrowed play themes vs. non-talking versions |
| Interactive storytelling chatbot | Kids aged 4–8 attributed strong agency to it, most at age 4 |
| Talking doll (moral-judgment study) | Shifted kids' moral judgments, comparable to a human participant |
| Smart speakers | Rated more factually capable than people; human names increased trust |
Source: Ying Xu, Lauren Girouard-Hallam, and Zhonghao Shi, cited via EdTech Innovation Hub.
What should parents watch for in an AI toy today?
None of this means AI toys are automatically harmful — it means design choices matter. The traits researchers keep flagging, like data collection, engineered emotional attachment, and unclear boundaries, are the same ones parents can check for at the shelf: does the toy have a camera or only a microphone and speaker, can a parent set topics and limits rather than just power it on and off, is it a one-time purchase or a subscription, and what certifications back its safety claims.
Ted, the interactive plush from Ted&Co, was designed around exactly these questions: no camera, real-time conversation in 60+ languages, and full parental control over topics and limits through the Ted&Co app. You can see the certifications and privacy details on the security and certifications page.
Harvard's findings are still to come. But the questions its researchers are asking, about trust, attachment, and what a toy actually says to a child, are worth asking at the shelf, not after the study wraps in a year or two. For more on how these toys process conversation, see our piece on how an AI toy actually understands what your child says.
Frequently asked questions
What is the Harvard Child-Centered AI Lab studying?
It's testing how children play with a purpose-built AI toy compared to traditional toys, tracking language use, social interaction, trust, and attachment, in a study that began in August 2026.
Is the Harvard AI toy study finished?
No. The lab is in the first, lab-based phase. A second, longer home-based phase is planned but not yet scheduled.
Do AI toys collect data on my child?
It depends on the toy. Some AI toys use cameras and store conversation data; others, like Ted, use only a microphone and speaker and are built without cameras, following GDPR and COPPA principles.
How can I tell if an AI toy is safe for my child?
Look for no camera, clear parental controls, a one-time price with no subscription, and recognized certifications (CE, FCC, GDPR, COPPA). See our FAQ for more.