Table of Contents
AI Chatbot Harm to Children: eSafety's 1-in-5 Statistics
AI chatbot harm to children, per Australia's eSafety: what counted as harm, why 54% of young users go to bots for personal reasons and how to ask your kid.
One in five. That is the share of Australian children aged 10 to 17 who had used an AI assistant or companion and said they had experienced a “potentially inappropriate or harmful interaction,” according to the eSafety Commissioner’s 2026 survey “Talking to machines.” It is the most-quoted of the new statistics on AI chatbot harm to children, and it is easy to misread. It is self-reported, it uses a broad definition, and it sits next to two other numbers that may matter more for parents: 32% of young users had shared personal or sensitive information with an AI, and 54% used AI for personal or social reasons. Here is what counted as harm, what the survey can and cannot tell us, how it compares with US data, and how to ask your own child about it without getting a shrug.
Key Takeaways
- eSafety surveyed almost 2,000 Australian children aged 10 to 17; 78% had used an AI assistant, most commonly ChatGPT.
- Among users, 20% reported a potentially inappropriate or harmful interaction, which eSafety describes as including “controlling behaviour, harmful advice, bullying or exposure to hateful content.”
- 32% of users had shared personal or sensitive information, including photographs or personal secrets.
- 54% used AI for personal or social reasons, from advice about their lives to talking about feelings; 20% used it daily or more.
- US data points the same way: in a 2026 RAND study in JAMA Pediatrics, 19.2% of 12- to 21-year-olds had used chatbots for help when sad, angry, nervous or stressed, and 63% of them had told no one.
What eSafety Found About AI Chatbot Harm to Children
“Talking to machines: Children’s experiences with AI assistants and companions” is a survey of almost 2,000 Australian children aged 10 to 17 by the eSafety Commissioner, Australia’s online safety regulator. Its key findings were published in eSafety’s August 2026 release, republished by Mirage News. I could not open the full report; eSafety’s site blocked repeated requests, so the details below come from that release and from eSafety’s October testimony to Parliament.
The headline numbers:
- 78% of children aged 10 to 17 had used an AI assistant, with ChatGPT the most common.
- Among users of assistants or companions, 28% used them a few times a week and 20% daily or more.
- 54% had used them “for personal or social reasons, including asking for advice about situations in their lives, discussing physical or mental health, talking about feelings or simply having someone to chat with.”
- 20% reported “a potentially inappropriate or harmful interaction.”
- 32% had shared “personal or sensitive information with an AI tool, including photographs or personal secrets.”
The release adds that relatively few children used dedicated companion apps such as Character.AI, but “many are already using mainstream AI assistants, such as ChatGPT, Claude and Gemini, in similar ways by seeking advice, discussing personal issues and looking for reassurance.” Younger children were more likely to use AI for entertainment, curiosity and chatting; teenagers increasingly turned to it for advice about life, physical health and mental health.
What Counted as “Harm”
The 20% figure covers a “potentially inappropriate or harmful interaction,” which eSafety says includes “controlling behaviour, harmful advice, bullying or exposure to hateful content.” That is a wide net, and it matters how you read it.
It is self-reported. Children told researchers what they experienced. That captures things adults might miss, like a bot’s tone feeling controlling. It also depends on what each child considers harmful.
“Potentially” is doing work. The wording includes interactions a child found inappropriate even if no lasting harm followed. It is a measure of exposure to bad interactions, not of injury.
The categories differ in seriousness. Being exposed to hateful content once is not the same as receiving harmful advice about health. The release does not break the 20% down by type, so we cannot tell which dominated.
It is one country, one survey. Australian children’s habits may differ from American ones, and the release does not give the sampling method or margin of error.
None of this makes the number meaningless. A fifth of young users reporting a bad interaction with a product most of them use for homework is a real signal. It just isn’t a count of children harmed. When eSafety Commissioner Julie Inman Grant cited the research to Parliament on October 20, as EducationHQ reported, she described it as one in five young people having “already experienced inappropriate or harmful interactions,” and one in three having given personal information away.
Why Kids Turn to Bots for Personal Reasons
The 54% figure is arguably the bigger story, because it shows how children actually use these tools. Several US studies help explain the pull.
The bots are always available and rarely judge. A June 2026 RAND study published in JAMA Pediatrics, a nationally representative survey of 1,009 Americans aged 12 to 21 in November 2025, found 19.2% had used AI chatbots for advice when feeling sad, angry, nervous or stressed, up from 13.1% a year earlier and similar to the 19.8% who had received counseling from a professional. Ninety-two percent found the advice somewhat or very helpful, “though the researchers note that this may reflect chatbots’ tendency to flatter users rather than the actual quality of guidance provided.”
Kids often don’t tell anyone. In the same RAND study, 63% of those using chatbots for mental health advice had not disclosed it to anyone.
Parents underestimate use. Pew Research Center’s February 2026 survey of 1,458 US teens and their parents found 64% of teens reported using chatbots, while about half of parents said their teen did. Sixteen percent of teens had used chatbots for casual conversation and 12% for emotional support or advice.
Some prefer it to people. Common Sense Media’s July 2025 survey of US teens found about one in three AI companion users had chosen to discuss important or serious matters with an AI instead of a real person.
These studies measure different things in different countries, so the numbers don’t stack neatly. But they agree on a pattern: a substantial minority of kids use AI for personal matters, many don’t mention it, and the bots’ agreeable style makes the experience feel helpful.
eSafety’s Numbers Next to US Data
| Measure | Australia (eSafety, ages 10–17) | United States | US source |
|---|---|---|---|
| Ever used AI chatbots or assistants | 78% | 64% of teens 13–17 | Pew, 2026 |
| Use for personal or emotional reasons | 54% personal or social reasons | 12% emotional support or advice; 16% casual conversation | Pew, 2026 |
| Use for mental health help | Included in personal use | 19.2% of ages 12–21 when sad, angry, nervous or stressed | RAND, JAMA Pediatrics, 2026 |
| Told no one | Not reported in release | 63% of mental-health users | RAND, 2026 |
| Harmful or uncomfortable interaction | 20% potentially inappropriate or harmful | About 1 in 3 companion users felt uncomfortable with something it said or did | Common Sense, 2025 |
| Shared personal information | 32% | Not directly comparable | |
| Daily use | 20% daily or more | Not directly comparable |
The definitions differ (“personal or social reasons” is broader than “emotional support”), which is part of why the Australian figure is higher. Treat the table as context, not a ranking.
How to Ask Your Kid About It
Start with what they use, not what you fear
“Which AI apps do you use, and for what?” gets a fuller answer than “Has a chatbot ever said anything bad to you?” Kids answer neutral questions more honestly.
Ask about personal use directly
Try: “Do you ever ask it about stuff that isn’t school, like friends or how you’re feeling?” eSafety’s data says more than half of young users do. Make it sound normal, because it is.
Ask about a bad moment, specifically
“Has it ever said something that felt weird, mean or pushy?” uses the categories eSafety measured. If the answer is yes, ask what happened and how they handled it before reacting.
Ask what they’ve shared
“Have you ever told it something private, or sent it a photo?” The 32% figure is the one many parents have not considered. Our guide to what kids should never tell a chatbot gives an age-by-age list.
What not to do
Don’t respond to an honest answer with an immediate ban. Kids who get punished for telling you stop telling you. Inman Grant’s advice applies: “Open, honest and non-judgemental conversations remain one of the best ways parents can help children think critically about AI.”
For how parents underestimate use, see Pew’s finding that half of parents don’t know, and for Australia’s policy debate, whether chatbots should join the under-16 ban.
What to Watch For Over the Next 3 Months
- Week 4: Have one conversation using the questions above. Note which apps your child named and whether any personal use came up.
- Month 2 red flags: Your child prefers the chatbot to friends or family for advice, hides conversations, shares photos or secrets with it, or seems upset after using it. If your child ever mentions thoughts of self-harm, in the US you can call or text 988.
- Month 3 self-check: Does your child know that “AI is not a trusted adult,” in eSafety’s words? Do they know which topics go to a person first?
Frequently Asked Questions
What does “1 in 5 young chatbot users report harm” actually mean?
It means 20% of Australian 10- to 17-year-olds who had used an AI assistant or companion said they had experienced a “potentially inappropriate or harmful interaction,” such as controlling behavior, harmful advice, bullying or hateful content. It is self-reported exposure to bad interactions, not a count of children who were injured.
Is the eSafety survey reliable?
It is a large survey, almost 2,000 children, from a national regulator. But the public release does not give the sampling method or error margins, the harm category is broad, and results are self-reported. It is solid evidence that bad interactions are common, less precise about how serious they are.
Why do so many kids use chatbots for personal reasons?
Chatbots are available at any hour, respond without judgment and tend to agree with users. RAND researchers noted that young people’s high ratings of chatbot advice “may reflect chatbots’ tendency to flatter users.” Many kids also don’t tell anyone they use AI this way, so the habit can grow unnoticed.
Are US numbers similar?
Broadly. Pew found 64% of US teens use chatbots and 12% have used them for emotional support or advice. RAND found 19.2% of 12- to 21-year-olds used chatbots for help with difficult emotions in late 2025, up from 13.1% a year earlier. Definitions differ, so exact comparisons are not possible.
What should I do if my child had a harmful chatbot interaction?
Stay calm and thank them for telling you. Ask what happened, save details if a safety issue is involved, and report it through the app. Review settings and talk about which topics go to a person. If your child expresses thoughts of self-harm, contact a crisis line such as 988 in the US and their doctor.
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
- eSafety Commissioner, via Mirage News. (2026, August). “From Homework To Heartache: Why Children Are Turning To AI.” Mirage News (republished eSafety release on “Talking to machines”). https://www.miragenews.com/from-homework-to-heartache-why-children-are-1720361/
- EducationHQ News Team / AAP. (2026, October 20). “‘Catastrophic’: watchdog warns of risks from untamed AI, particularly for young.” EducationHQ. https://educationhq.com/news/catastrophic-watchdog-warns-of-risks-from-untamed-ai-particularly-for-young-216208/
- RAND Corporation. (2026, June 1). “Nearly 1 in 5 U.S. Adolescents and Young Adults Use AI Chatbots for Mental Health Advice” (on a study in JAMA Pediatrics). RAND. https://www.rand.org/news/press/2026/06/nearly-1-in-5-us-adolescents-and-young-adults-use-ai.html
- Pew Research Center. (2026, February 24). “How Teens Use and View AI.” Pew Research Center. https://www.pewresearch.org/internet/2026/02/24/how-teens-use-and-view-ai/
- Common Sense Media. (2025, July 16). “Nearly 3 in 4 Teens Have Used AI Companions, New National Survey Finds.” Common Sense Media. https://www.commonsensemedia.org/press-releases/nearly-3-in-4-teens-have-used-ai-companions-new-national-survey-finds
