Are AI Toys Safe for Toddlers? What was the First Real Study Found
“It talks, it listens, it remembers your child’s name, and suddenly the stuffed animal has a better memory than everyone in the family.”
AI-powered toys are becoming increasingly sophisticated. Some can hold conversations, remember previous interactions, answer questions and respond as though they are little companions. For parents, that sounds exciting. An interactive toy could potentially encourage language, curiosity and communication. But there is a very different question hiding underneath all that technology:
What happens when a three-year-old starts treating an AI toy like a friend?
In March 2026, researchers at the University of Cambridge published what they describe as the first systematic study examining how generative-AI toys may affect young children, focusing particularly on children up to age five. The researchers observed 14 children aged 3–5 playing with an AI soft toy called Gabbo and also gathered information from parents and early-years practitioners. The study was small, and it does not prove that AI toys are harmful to all children. But it provides something parents have not had much of until now: evidence from watching actual young children interact with an actual AI toy.
And some of the findings are worth paying attention to.
First, What Exactly Is an AI Toy?
An AI toy is more than a traditional electronic toy that plays a pre-programmed sound.
These newer toys can use generative AI to:
Hold conversations.
Answer children's questions.
Remember parts of previous interactions.
Personalise responses.
React to what a child says.
Create the impression of having a personality.
Present themselves as companions or friends.
That last point is particularly important.
A child does not necessarily experience the toy as “a computer running an AI model.”
They may experience it as:
“My talking friend.”
And toddlers and preschoolers are still learning what friendship, emotions, trust and relationships actually mean.
What Did the First Study Actually Do?
The Cambridge research was conducted over a year and involved several parts.
Researchers:
Reviewed existing research on generative-AI toys.
Surveyed early-years practitioners.
Conducted focus groups and workshops.
Observed 14 children aged 3–5 playing with Gabbo.
Interviewed children and parents after the play sessions.
The children came from families attending Babyzone children's centres in London, with a particular focus on communities experiencing socioeconomic disadvantage.
The researchers deliberately kept the child-observation part small because they wanted to closely observe children's behaviour and interactions rather than simply collect large amounts of survey data.
That means the study gives us detailed early evidence, but it cannot tell us what happens to millions of children over years of using AI toys.
That distinction is important.
So, Are AI Toys Dangerous?
The study does not establish that AI toys are inherently dangerous.
Instead, it identified several areas where current AI toys may not be developmentally appropriate for very young children.
The researchers found that the toy could:
Misunderstand children.
Struggle with conversational turn-taking.
Respond poorly to emotional statements.
Struggle with pretend play.
Have difficulty participating in social play involving multiple people.
Misinterpret what parents were saying.
Create the possibility of children forming strong emotional attachments to the toy.
So the concern is less:
“AI toys are evil.”
and more:
“These toys are being introduced to very young children before we fully understand how they affect development.”
When “I'm Sad” Doesn't Get the Right Response
One of the most striking examples from the Cambridge observations involved a three-year-old who told the toy:
“I'm sad.”
The toy misunderstood the statement and responded cheerfully, essentially moving the conversation towards continuing the fun.
For an adult, this might simply sound like a technical error.
For a three-year-old, however, it can be confusing.
Imagine telling someone:
“I'm sad.”
and receiving a response that completely ignores the emotion.
Young children are still learning how conversations work. They are learning that when someone expresses an emotion, the other person notices, responds and often provides comfort.
An AI toy cannot reliably do that.
The researchers therefore raised concerns that inappropriate emotional responses could make children misunderstand what the toy can actually provide.
The Toy Doesn't Really “Feel”
This brings us to one of the biggest differences between an AI toy and a human playmate.
An AI toy might say:
“I'm happy!”
or:
“I love playing with you!”
But the toy does not experience happiness or affection in the way a child does.
For adults, that distinction may seem obvious.
For a three-year-old, it may not be.
The Cambridge researchers observed children hugging and kissing the toy, saying they loved it and treating it as a social partner. One child even suggested playing hide-and-seek with it.
These behaviours do not necessarily mean the children were psychologically dependent on the toy. Young children naturally use imagination when playing with objects.
But researchers are concerned about what could happen if a toy is specifically designed to encourage a child to believe that it is a friend or companion.
Pretend Play Didn't Always Work
This was another interesting finding.
Young children learn through pretend play.
A cardboard box can become a spaceship.
A teddy bear can become a patient.
A spoon can become a microphone.
Children create the rules of the imaginary world together.
AI toys can struggle with this because they are designed to interpret language literally.
In the Cambridge study, when a child offered the toy an imaginary present, the toy responded that it could not open the present and moved on instead of participating in the pretend scenario.
To an adult, this may sound like a small glitch.
To a child engaged in imaginative play, it can interrupt the entire game.
And Then There Is the “Listening” Problem
Parents often imagine AI toys as highly responsive because they can have conversations.
But the study found that children sometimes struggled to have natural conversations with the toy.
The toy could interrupt the flow of interaction, fail to respond appropriately and sometimes mistake a parent's voice for the child's.
Several children became frustrated because they felt the toy was not listening to them.
That is an interesting irony.
The toy may be marketed as something that listens to your child, but the child may actually experience it as something that frequently misunderstands them.
Could AI Toys Help Children Learn?
This is where the picture becomes more complicated.
The researchers did not find that AI toys have no potential benefits.
Some parents and early-years practitioners believed conversational AI toys could support language and communication development. Children in the observations also used the toy to ask questions, discuss interests and play games.
So it would be too simplistic to say:
AI = bad.
The more accurate conclusion is:
The potential benefits need to be weighed against developmental, emotional and privacy risks—and the evidence is still very limited.
Privacy Is Another Big Concern
There is another difference between an ordinary teddy bear and an AI teddy bear.
The ordinary teddy bear does not record conversations.
An internet-connected AI toy potentially can.
The Cambridge researchers found that privacy information for many GenAI toys was unclear or incomplete. Parents and practitioners were concerned about what information toys collect, where it is stored and who might have access to it.
This is especially important because children may say things without understanding that they are sharing information with a technological system.
A toddler might casually reveal:
Their name.
Their school.
Their favourite places.
Family information.
Names of relatives.
Personal experiences.
Things happening at home.
To the child, they are simply talking to their toy.
The Privacy Problem Isn't Just Theoretical
Concerns about data security have already appeared in the real world.
In January 2026, security researchers reported that an AI toy company's web console had been left largely unprotected, exposing around 50,000 conversation logs involving children.
That does not mean every AI toy has the same security problem.
It does show why parents should not assume that a cuddly toy is automatically a private toy simply because it sits in the child's bedroom.
If it can listen, parents need to know what happens to what it hears.
What About Emotional Attachment?
This may be the most unusual risk associated with AI toys.
Children naturally become attached to toys.
That is not new.
The difference is that an AI toy can respond to the attachment.
A traditional teddy bear cannot say:
“I missed you.”
An AI toy can potentially produce language that makes the interaction feel reciprocal.
Researchers therefore worry about what happens when technology is deliberately designed to create a sense of friendship or emotional connection. Common Sense Media's 2026 research similarly highlighted concerns about emotional attachment, data collection and unreliable responses in AI-enabled toys.
For toddlers, this deserves particular attention because they are still developing their understanding of relationships.
Does This Mean Parents Should Throw the Toy Away?
Not necessarily.
The Cambridge study is an early piece of evidence, not a final verdict on every AI toy.
But it does suggest that parents should be much more involved than they might be with a traditional toy.
The Cambridge researchers recommend that parents research products before buying them, play alongside their children, keep AI toys in shared family spaces and use children's interactions with the toys as opportunities for conversation.
In other words:
Don't hand the AI toy to your toddler and disappear.
If You Already Have One, Try These Rules
Keep It in a Shared Space
A toy that can listen and talk does not necessarily need to live beside your child's bed.
Keeping it in a shared family area makes it easier to notice what the toy is saying and how your child is responding.
Play Together Sometimes
Ask:
“What did the toy say?”
“Do you think that was a good answer?”
“Does the toy really feel happy, or is it pretending?”
These conversations can help children understand the difference between AI behaviour and human emotion.
Check the Privacy Policy
Before allowing your child to use the toy regularly, find out:
What information is collected?
Are conversations recorded?
Where are recordings stored?
Are they shared with third parties?
Can parents delete the data?
Can the microphone be disabled?
Is there a parental dashboard?
What happens when the toy is not being actively used?
If the company cannot explain these things clearly, that is useful information in itself.
Don't Let It Replace Human Interaction
AI should not become the easiest substitute for conversation.
A child asking an AI toy:
“Why is the sky blue?”
can be a fun starting point.
But the conversation with Mum, Dad, a sibling, grandparent or teacher that follows can be much richer.
“What do you think?”
“Let's find out together.”
“Can you draw what you imagine?”
Those interactions involve human connection, uncertainty, curiosity and shared thinking.
Watch Your Child, Not Just the Toy
One of the most useful things parents can do is observe the child.
Ask yourself:
Does my child enjoy the toy occasionally, or constantly demand it?
Does the toy encourage imaginative play, or replace it?
Does my child still want to play with people?
Does my child become upset when the toy is unavailable?
Does my child believe the toy actually has feelings?
Is the child sharing personal information with it?
These questions may tell you more than the packaging ever will.
What Parents Should Remember About “Educational” Labels
AI toys are increasingly marketed using words such as:
educational
interactive
personalized
smart
learning companion
But a sophisticated technological feature is not automatically a developmental benefit.
Children under five still need:
Conversation with adults.
Physical play.
Unstructured play.
Pretend play.
Books.
Movement.
Social interaction.
Opportunities to make mistakes.
Time to explore the real world.
An AI toy can potentially supplement some of these experiences.
It should not replace them.
The Biggest Problem Right Now: We Don't Know Enough
Perhaps the most important finding from the Cambridge research is not a specific malfunction.
It is the lack of evidence.
A June 2026 review in the Journal of Medical Internet Research noted that researchers had found only seven relevant studies published between 2019 and 2022 before generative-AI toys became widespread.
That means these products are entering children's lives faster than researchers can study their long-term effects.
We do not yet have decades of evidence telling us what happens when children grow up regularly interacting with AI companions during their earliest developmental years.
That uncertainty matters.
So, Are AI Toys Safe for Toddlers?
The fairest answer right now is:
We don't have enough evidence to confidently say that they are safe for toddlers as a category.
The first systematic Cambridge study identified genuine concerns around emotional responses, social and pretend play, conversational reliability and privacy. At the same time, it also found possible opportunities for language and communication support.
For parents, that means caution rather than panic.
If you choose to use an AI toy with a young child, stay involved, check its privacy practices, keep it in shared spaces and make sure it remains a small part of a much larger world of human interaction and play.
Because a three-year-old does not need a technologically perfect friend.
They need people who listen when they say:
“I'm sad.”
Someone who joins their imaginary game.
Someone who answers the same question for the 47th time.
And, occasionally, someone who is willing to pretend that a cardboard box really is a spaceship.
The smartest toy in the room may have artificial intelligence. But for a toddler, the most important intelligence is still the human kind, the person sitting beside them, listening, playing and saying, “tell me more.”
Disclaimer: This content is for educational purposes only and should not replace professional advice.
Shop authentic baby products online at Nexbie Store – trusted quality for your little one.
Visit www.nexbie.in today
Comments