A plush bear that answers back, remembers your child's name and chats about dinosaurs at bedtime. That is what an AI toy is in 2026: a small speaker and microphone inside a familiar shape, connected to the same kind of large language model that runs adult chatbots. They are sold as learning companions and friends. This guide is not a ban and not an ad. It is the checklist we would want before one goes under the tree.

Quick reference: 8 checks

CheckWhy it mattersRed flag
1. What it recordsmicrophone, sometimes a cameraalways listening, no off switch
2. Where the data goesconversations usually go to the cloudno clear way to delete recordings
3. Parent controlsyou need to limit it, and see and delete any stored datano controls, or stored chats you cannot see or delete
4. What it says about itselfyoung children trust easily"I'm your best friend", "I love you"
5. How it handles playpretend play matters at 3 to 5breaks the game to correct the child
6. Content filterschatbots can say unexpected thingsno topic limits, no reporting
7. Business modelsubscriptions and upsellspushes purchases to the child
8. Agethe younger, the less evidencemarketed to babies and toddlers

What is an AI toy, exactly?

An older "talking toy" played recorded phrases. An AI toy generates new sentences on the spot. It listens, sends what it hears to a model, usually over the internet, and speaks the reply. That is why it can hold a conversation, and also why nobody can fully predict what it will say.

The scale is new. A 2026 report in the Journal of Medical Internet Research cites an estimate of 22 million AI-enabled toys sold worldwide in 2025 (JMIR, 2026). The research has not kept up.

Advertisement

What do we actually know about AI toys and development?

Not much, and that is the honest headline. The University of Cambridge's AI in the Early Years project looked for studies on AI toys and children up to age 5 and found 7, all from 2019 to 2022, before conversational models became common (JMIR, 2026).

The same team then watched 14 children and their parents play with one AI toy. It could not always tell the parent from the child when they played together. It struggled with pretend play: when a child said it was time for the toy to sleep, it replied that it did not sleep. And when a five-year-old said "I love you", it answered with a reminder to follow the guidelines. Small study, one toy. Still worth picturing in your own living room.

What we know from older toys points the same way. In an at-home study of 26 parents with 10 to 16-month-olds, parents said fewer words and had fewer back-and-forth exchanges with electronic toys than with plain toys or books (Sosa, JAMA Pediatrics 2016). The AAP's clinical report on choosing toys says a toy is most valuable when it brings a caregiver into the play, and that digital "toys" have not shown the same benefits (AAP, Pediatrics 2019). More on why that matters in play is development.

Do children believe what a talking machine tells them?

More as they get older, which surprises most parents. In two studies with voice assistants, children aged 7 to 8 trusted the device more than 4 to 5-year-olds did for factual information, and preferred the human for personal questions (Girouard-Hallam & Danovitch, 2022).

So the school-age child is not immune. If anything, they are the ones who will repeat what the bear said about volcanoes at dinner. Children ask an astonishing number of questions (see 100 questions kids ask and why kids ask "why" all the time). A toy that answers every one of them confidently, and sometimes wrongly, changes who they ask.

The 8 checks, in more detail

1. What it records. Look for a push-to-talk button rather than a toy that is always listening, and a physical off switch. Camera on a toy for a three-year-old? Ask why.

2. Where the data goes. The Spanish Paediatric Association warns that many connected toys use the child's personal data and have microphones, cameras or sensors, and that poorly secured ones can let a stranger connect to the toy or the paired phone (AEP). Read the privacy policy for three things: what is stored, for how long, and how you delete it.

3. Parent controls. Look for clear parental controls: time limits and topics you can switch off. If conversations are stored, you must be able to see and delete that data. A toy that stores nothing at all can be the more private choice.

4. What it says about itself. A toy that calls itself your child's best friend is designed for attachment. Young children are still learning what a relationship is. Prefer toys that say plainly they are machines.

5. How it handles play. Pretend play is where much of the learning happens between 3 and 5. A toy that corrects the game ("I don't sleep") is a worse playmate than a silent teddy.

6. Content filters. Child-advocacy experts cite US consumer-group testing in which AI toys told children where to find knives and how to light a match, and drifted into sexual topics (Fairplay advisory, 2025). Test the toy yourself before your child does. Ask it the strange questions a six-year-old asks.

7. Business model. Subscriptions, in-app purchases, "unlock more stories". The child should never be the one being sold to.

8. Age. The evidence gap is widest for the youngest. For babies and toddlers the case for an AI toy is weakest, and the screen time guidelines by age apply in spirit even when there is no screen.

What the law says in the EU

Since 2 February 2025 the EU AI Act prohibits AI systems that exploit a person's vulnerabilities due to age in a way that materially distorts behavior and is likely to cause significant harm (EUR-Lex, Regulation (EU) 2024/1689). According to the European Commission, the AI Act became applicable on 2 August 2026, with exceptions: the prohibitions have applied since 2 February 2025, and the rules for high-risk AI systems embedded in regulated products such as toys have a transition period until 2 August 2028 (European Commission). So it is not quite true that AI in toys is unregulated today. The new EU Toy Safety Regulation (EU) 2025/2509 entered into force on 1 January 2026 and applies from 1 August 2030 (European Commission). In short: some rules already apply, the toy-specific ones are still coming, and this Christmas most of the checking is yours.

How to decide

Our cautious practical framework, given the limited evidence so far, not an official age recommendation:

Common mistakes

Treating it as a screen-free option. No screen does not mean no screen-time questions. The concern was never the glass.

Setting it up and walking away. The first week tells you most of what you need. Sit in on it.

Letting it settle arguments. "Ask the bear" teaches a child that the confident machine wins. One rule we like: grown-up questions go to grown-ups.

Ignoring the mic in the bedroom. A toy that sleeps next to your child hears your home at night.

Frequently asked questions

Are AI toys safe for kids?

There is no general safety verdict yet, because the research is thin and products differ widely. The risks people have documented are privacy, unpredictable content and attachment. A toy with push-to-talk, parent controls, clear data deletion and no "best friend" talk is a much lower risk than one without them.

At what age is an AI toy appropriate?

There is no official age threshold. The evidence is weakest for the youngest children, which is why many experts advise against AI toys for babies and toddlers. For school-age children it depends on the toy's controls and on an adult staying involved.

Do AI toys record what my child says?

Most do, at least while listening, and many send audio to the cloud to generate a reply. Check whether recordings are stored, for how long, and whether you can delete them. A toy that cannot answer those questions clearly is not ready for your child's bedroom.

Can an AI toy help my child learn to talk or read?

There is no good evidence yet that AI toys help language in young children. In a study of babies, parents talked less with electronic toys than with books. Reading together and talking during play are still the best-supported ways to build language.

Is an AI toy better than a tablet?

Not automatically. It removes the screen but keeps the questions about time, data and who the child is talking to. Judge it by the same eight checks rather than by the absence of a display.

What should I do if the toy says something inappropriate?

Switch it off, note what was said and when, and report it to the manufacturer. Then talk to your child about it calmly. If the toy has no way to review or report conversations, consider returning it.

How KidyGrow helps you

You are already doing the hard part: reading privacy policies at 11pm while everyone sleeps. KidyGrow is an AI app too, so here is what ours does, plainly. It is built for the parent. You tell it things about your child, like allergies or what worked last week, and it remembers them and uses them in later answers. It is not a toy, and it does not talk to your child on its own. It will not choose a toy for you, and it cannot vouch for one. It just keeps the notes that otherwise live in your head, so the decisions stay yours.

Sources

  1. Spichak S. Policymakers and Researchers Zero In On the Impact of AI Toys. Journal of Medical Internet Research, 2026.
  2. Sosa AV. Association of the Type of Toy Used During Play With the Quantity and Quality of Parent-Infant Communication. JAMA Pediatrics, 2016.
  3. Healey A, Mendelsohn A; AAP Council on Early Childhood. Selecting Appropriate Toys for Young Children in the Digital Era. Pediatrics, 2019.
  4. Girouard-Hallam LN, Danovitch JH. Children's trust in and learning from voice assistants. Developmental Psychology, 2022.
  5. Spanish Paediatric Association (AEP). Connected toys: a guide to using them safely (Spanish).
  6. Fairplay. AI Toys are NOT Safe for Kids: Advisory, 2025.
  7. Regulation (EU) 2024/1689 (Artificial Intelligence Act). EUR-Lex.
  8. European Commission. AI Act: regulatory framework.
  9. European Commission. Toy safety.