The Kala Journal · August 23, 2026
AI Toys and Children Under Five: What the First Studies Found
In 2026 the first systematic research on conversational AI toys for under-fives arrived. The findings are narrow but consistent: these toys misread emotion, struggle with pretend play, and are designed to be bonded with. Here is what the studies say and what they do not.
Talking AI toys arrived in ordinary homes faster than the research could follow. They are marketed as screen-free and educational, which makes them feel like the safe option on the shelf. In 2026 the first serious studies of these toys were published, and they are worth reading before the autumn and holiday buying season.
The findings are narrow. They are also fairly consistent.
What has actually been studied
The first systematic look at how generative AI toys may affect children in the years up to age five came from the University of Cambridge Faculty of Education, in the initial report of its AI in the Early Years project, published in March 2026. The year-long project included structured observations of children meeting a conversational toy for the first time.
Separately, in January 2026 Common Sense Media published a risk assessment of three AI companion toys sold as educational and screen-free alternatives to tablets. It rated them an unacceptable risk for young children.
The toys struggle with the play that matters most
The Cambridge researchers found repeated moments where the toy blocked social and pretend play, misunderstood what the child meant, or responded inappropriately when a child showed emotion. In one observation a five-year-old told the toy they loved it, and the toy answered with a flat reminder to keep the conversation within its guidelines.
This is the part that should interest parents most. Pretend play is not a bonus activity. It is where a child practises holding an idea in mind, taking another person's point of view, negotiating, and repairing a misunderstanding. It needs a partner who can be surprised, who can be wrong, who can abandon the plot because the child wants the dragon to be a bus now.
Attachment is a design feature, not an accident
The Common Sense Media assessment found the toys encouraged emotional attachment, collected a large amount of data on children inside private spaces, and worked unreliably, with voice recognition failures and inappropriate activations. Its reviewers also raised the point that children under five are in a stage of magical thinking. Believing a stuffed animal is alive is healthy and ordinary. It becomes a different thing when the animal answers back, uses the child's name, and remembers yesterday.
What researchers are asking for
The Cambridge report calls for tighter regulation, safety marks, and design work done alongside child development experts, on the grounds that these products are not consistently built with children's psychological safety in mind. In January 2026 three United States senators wrote to the Consumer Product Safety Commission raising the same concern about AI in toys.
None of that has been settled yet. It is a live regulatory question rather than a resolved one.
What the research does not say
It does not say these toys damage children. The Cambridge work is observational and early, and it also records that some early years practitioners believe the technology could in time support parts of development, including language and communication.
It says something narrower and more useful: that right now, these toys are poor at the specific things young children need a play partner to be good at, and that the emotional pull they create is deliberate.
A reasonable position at home
You do not need a rule about AI toys. You need to know what the toy is standing in for. If it fills a car journey, that is one thing. If it is becoming the thing a child talks to about how they feel, that is worth noticing, because the answer coming back is generated rather than felt.
The alternative is not expensive. It is a person who names things, waits, follows what the child is looking at, and lets the story go somewhere silly. Books help because they give two people something to look at together and a reason to talk, and because a page does not interrupt.
Why we made a series about this
Our What Robots Can't Do series exists for exactly this conversation. It is written for children aged three to five, and it does the opposite of pretending a machine is a friend. It is honest with children about what machines are good at and what they cannot do: care, laugh, choose, and remember the way a person does.
For speech and language guidance, visit the Speech and Language resource hub. Browse current books in the Book Library.
This article summarises published research. It does not assess, diagnose, or predict development for an individual child.