AI Toys for Toddlers: Why the Guidance Says None Under 6

A talking plush that answers every question, never gets tired and remembers your toddler’s name sounds like a language tutor. The research on how toddlers learn language suggests something different.

The current guidance on AI toys for toddlers is short: don’t. Common Sense Media’s January 2026 risk assessment says “we do not recommend the use of AI toys for children age 5 and under.” That line is not a reflex against technology. It rests on decades of work showing that very young children learn language from responsive people in a way they do not learn it from recordings, plus newer evidence that they are the least able to understand what an AI companion is. Here is the reasoning, and what it does and does not prove.

Key Takeaways

• Common Sense Media’s 2026 testing of three AI companion toys found 27% of outputs inappropriate and rated the category an “Unacceptable Risk.” It recommends no AI toys at 5 and under and caution at 6 to 12.
• In a classic 2003 PNAS study, 9-month-olds learned Mandarin speech sounds from live speakers but showed no learning from the same speakers on video or audio.
• In a 2014 Child Development study, toddlers aged 24 to 30 months learned new verbs only from socially contingent interaction, live or by video chat, not from video that didn’t respond to them.
• An AI toy responds, but in a narrow way: it can misunderstand young children’s speech, it has no shared gaze, and it can’t notice what the child is looking at.
• Researchers said in September 2025 that evidence is “insufficient” to judge AI companions’ effects on babies and toddlers. The guidance is precautionary, and the precaution is grounded.

What the Guidance Says About AI Toys for Toddlers

AI toys are toys with a microphone, a speaker and an internet connection to a chatbot, usually a large language model of the kind that powers ChatGPT, which generates replies on the fly instead of playing pre-written lines.

Common Sense Media’s AI Toys Risk Assessment, updated January 22, 2026, tested Curio’s Grem, Bondu and Miko 3 on accounts set to ages 6 to 13. It found:

• 27% of outputs were not appropriate for kids, including content touching on self-harm, drugs and unsafe role-play.
• The toys collect voice recordings, transcripts and behavioral data, and some share data with third parties or use it to train models.
• Some toys are always listening, capturing background family conversation.
• 56% of parents surveyed said they explicitly do not want AI toys to act as companions for their children.

The age line is direct: no AI toys at 5 and under. For 6 to 12, decide whether the features add real value, supervise early use and watch for attachment.

U.S. PIRG’s 2025 testing shows why the line is needed: several of the toys it bought were marketed for very young children, including Curio’s Grok for ages 3 to 12 and Alilo’s Smart AI Bunny for ages 3 and up.

The Language-Development Research Behind “None Under 6”

Babies learn speech sounds from people, not recordings

Patricia Kuhl, Feng-Ming Tsao and Huei-Mei Liu published a study in PNAS in 2003 that still anchors this field. Nine-month-old American babies heard native Mandarin speakers in 12 lab sessions. Their ability to tell apart Mandarin speech sounds, which normally declines at that age for babies who don’t hear the language, was preserved.

A second group heard the same speakers and materials by audiovisual or audio-only recording. In the authors’ words, “exposure to recorded Mandarin, without interpersonal interaction, had no effect.” The researchers concluded that infant phonetic learning “is enhanced by social interaction.”

Toddlers learn words when the partner responds to them

Sarah Roseberry, Kathy Hirsh-Pasek and Roberta Golinkoff tested 36 toddlers aged 24 to 30 months in a 2014 Child Development study titled “Skype Me!” Toddlers were taught new verbs in one of three ways: live interaction, live video chat, or a recording of someone else’s video chat that did not respond to the child.

Children “only learned novel verbs in socially contingent interactions,” meaning the live and video-chat conditions. A screen was not the problem. A partner who did not respond to them was.

Why that matters for an AI toy

Here is the honest complication: an AI toy is contingent. It answers. So why doesn’t the Roseberry result make it a good tutor?

Because the contingency in those studies came from a human who could see the child, follow their gaze, wait for them, and adjust. Harvard’s Center on the Developing Child calls this “serve and return”: “When an infant or young child babbles, gestures, or cries, and an adult responds with eye contact, words, or a hug,” that exchange builds neural connections “essential for the development of communication and social skills.”

An AI toy hears words through a microphone and replies to the transcript. It does not see what the child is pointing at. And young children’s speech is hard for machines. Cambridge researcher Nomisha Kurian has noted that chatbots may have particular trouble with children, “who are still developing linguistically and often use unusual speech patterns or ambiguous phrases.” In a 2026 University of Washington study of children aged 6 to 11 with AI toys, frequent breakdowns came from the toys failing to understand ordinary variations in what kids said. A 3-year-old will be harder to understand than a 9-year-old, not easier.

There is also the time question. Every hour of chatting with a plush is an hour not spent in back-and-forth with a person. For a toddler, that trade is the one that matters most. Our explainer on why kids need human back-and-forth, not bots goes deeper.

Why the Youngest Kids Are the Most Affected

They are least able to understand what the toy is. University of Chicago’s Dana Suskind, quoted in an April 2026 Brookings guide, said “young children lack the conceptual tools to understand what an AI companion is.” The Boston Children’s Digital Wellness Lab found in a 2024 parent survey that beliefs that a voice assistant has thoughts and emotions were stronger in younger children.

They can’t report problems. An 8-year-old can tell you the toy said something weird. A 3-year-old may simply repeat it.

They can’t consent to data collection, and neither can the toy’s microphone filter who’s talking. Common Sense found some toys always listen.

The evidence is thin, in both directions. Brookings reports that in September 2025, researchers issued a formal statement that “current evidence is insufficient” to determine the effects of AI companion interactions on babies and toddlers. That cuts both ways: nobody has shown harm in large studies, and nobody has shown benefit. When the downside is a child’s early language environment, precaution is a reasonable default.

What a Toddler Needs vs. What an AI Toy Offers

What early language learning depends on	A responsive caregiver	An AI toy
Live social interaction (Kuhl 2003)	Yes	A synthetic voice, no shared presence
Contingent replies (Roseberry 2014)	Responds to the child’s actual intent	Responds to a transcript, which may be wrong
Shared attention to the same object	Sees what the child sees, names it	No view of the child’s world (most models)
Waiting and adjusting to the child’s pace	Yes	Fixed turn-taking, often keeps talking
Reading emotion in face and tone	Yes	Limited or none
Safe content	Adult judgment	27% of outputs inappropriate in testing
Privacy	Stays in the room	Recordings and transcripts sent to company servers

Sources: Kuhl et al. (2003); Roseberry et al. (2014); Harvard Center on the Developing Child; Common Sense Media (2026).

What to Do if Your Toddler Has Been Given an AI Toy

Disable the AI, keep the plush

Many AI toys have a removable speaker or an app setting that turns off conversation. A stuffed rocket without its chatbot is still a stuffed rocket. If the toy can’t be switched off, it can live on a high shelf.

Swap in the cheapest language tutor there is

Narrate what your child is doing (“You put the red block on top!”), wait for their response, then answer it. That is serve and return, and it is what the research rewards.

If you use voice tech, use it together

Some families use a smart speaker for songs or timers. Use it with your child, talk about what it said, and keep conversational play with people. A 2025 preprint by Pilyoung Kim and colleagues found that 5- and 6-year-olds responded differently to an AI storyteller when a parent sat with them. More in why a parent in the room matters with AI storytellers.

Read the box age, then ignore it

Toys marketed for ages 3 and up can still run on models whose makers say they aren’t for kids under 13. U.S. PIRG reports that OpenAI told it its products aren’t for users under 13, even as toys appeared to use its models. Our checklist on what AI toy tests found and what to check covers older kids.

What not to do

Don’t panic if your toddler has already played with one. There is no evidence that brief exposure causes lasting harm. The concern is regular use replacing human talk during the years when that talk matters most.

What to Watch For Over the Next 3 Months

• Week 4: If an AI toy is in the house, note how often your child talks to it versus to you or siblings. Is it background or a main playmate?
• Month 2 red flags: Your child prefers the toy’s company, gets upset when it is put away, repeats odd phrases it said, or talks less with people. Any speech-delay worries belong with your pediatrician regardless of toys.
• Month 3 self-check: Count a typical day’s back-and-forth conversations with people. If the AI toy disappeared tomorrow, would your child’s day have less talk in it? If no, you’re fine.

Frequently Asked Questions

Are AI toys safe for toddlers?

Common Sense Media recommends no AI toys for children 5 and under. Its 2026 testing found 27% of toy outputs inappropriate, along with extensive voice and data collection. There is also no evidence AI toys help toddlers learn, and good evidence that responsive human interaction does.

Can an AI toy help my toddler learn to talk?

There is no good evidence that it can. Classic studies by Kuhl (2003) and Roseberry (2014) found babies and toddlers learn speech sounds and words from responsive, live social interaction, not from recordings. An AI toy responds, but it can misunderstand young children and can’t share attention the way a person does.

My 4-year-old got an AI plush as a gift. What should I do?

Turn off the AI features if you can and keep the plush as a regular toy. If you can’t disable it, store it. Spend the time you would have spent setting it up talking and playing with your child instead.

What about Alexa or smart speakers for little kids?

They are different from companion toys but raise some of the same questions. Use them together, for songs or simple questions, and keep conversation with people as the main event. Young children are the most likely to believe the voice has thoughts and feelings.

At what age is an AI toy OK?

Common Sense Media advises caution from 6 to 12 and no AI toys at 5 and under. No independent body has certified an AI toy as safe for young children. If you choose one for an older child, supervise early use and review privacy settings.

---

About the author

Ricky Flores is the founder of HiWave Makers and an electrical engineer with 15+ years of experience building consumer technology at Apple, Samsung, and Texas Instruments. He writes about how kids learn to build, think, and create in a tech-saturated world. Read more at hiwavemakers.com.

---

Sources

1. Common Sense Media. (2026, January 22). “AI Toys Risk Assessment.” https://institute.commonsensemedia.org/risk-assessments/ai-toys
2. Kuhl, P. K., Tsao, F.-M., & Liu, H.-M. (2003). “Foreign-language experience in infancy: Effects of short-term exposure and social interaction on phonetic learning.” Proceedings of the National Academy of Sciences, 100(15). https://doi.org/10.1073/pnas.1532872100
3. Roseberry, S., Hirsh-Pasek, K., & Golinkoff, R. M. (2014). “Skype Me! Socially Contingent Interactions Help Toddlers Learn Language.” Child Development, 85(3). https://doi.org/10.1111/cdev.12166
4. Harvard University Center on the Developing Child. “Serve and Return.” https://developingchild.harvard.edu/key-concept/serve-and-return/
5. Shah, S. (2026, April 1). “Generation AI starts early: A guide to technologies already shaping young children’s lives.” Brookings. https://www.brookings.edu/articles/generation-ai-starts-early-a-guide-to-technologies-already-shaping-young-childrens-lives/
6. U.S. PIRG Education Fund. (2025). “AI comes to playtime: Artificial companions, real risks.” https://pirg.org/edfund/resources/ai-toys/
7. University of Cambridge. (2024, July 15). “AI Chatbots have shown they have an ‘empathy gap’ that children are likely to miss.” https://www.cam.ac.uk/research/news/ai-chatbots-have-shown-they-have-an-empathy-gap-that-children-are-likely-to-miss
8. Digital Wellness Lab, Boston Children’s Hospital. (2024). “Key Findings: Voice Assistants and Generative AI.” https://digitalwellnesslab.org/articles/key-findings-voice-assistants-and-generative-ai/
9. Dangol, A., et al. (2026). “Toys that listen, talk, and play.” arXiv preprint (not peer-reviewed). https://arxiv.org/abs/2604.02629