The Chatbot Empathy Gap: Why Kids Miss That Bots Don't Care

Ask a chatbot “I failed my test and I feel stupid,” and you will likely get something warm: “I’m really sorry, that sounds so frustrating. You’re not stupid.” It reads like empathy. The chatbot empathy gap is the space between that sentence and what a person who cares would actually do next. Cambridge researcher Nomisha Kurian named it in a 2024 study and argued that children are the users most likely to miss it.

That matters because kids are not just using chatbots for homework. In Common Sense Media’s 2026 census, 37% of AI users aged 9 to 17 had discussed feelings or personal problems with AI, and a quarter of those said they sometimes feel AI “understands them better than most people do.” This article explains why warm-sounding bots earn so much trust, what is going on under the hood, and gives you activities to teach the difference at every age.

Key Takeaways

• The empathy gap, a term from Cambridge’s Nomisha Kurian, describes chatbots that mimic caring language without understanding a child’s emotional needs.
• In a University of Edinburgh study of 166 children aged 6 to 11, about two in three thought smart speakers might think like a human, and 70% said it is wrong to be rude to them.
• Chatbots produce empathic-sounding text by predicting what comforting replies usually look like, and they are tuned to agree, which makes them feel more understanding than they are.
• Real empathy includes stakes and follow-through: remembering, noticing, checking in, telling you something hard. Bots lack those.
• Kids can learn this with simple activities, from a “kindness jar” puppet for 5-year-olds to an empathy audit for teens.

What the chatbot empathy gap is

An empathy gap, in Kurian’s sense, is the mismatch between a chatbot’s human-like, caring style and its actual inability to understand or respond to a child’s emotional needs. Her paper, “‘No, Alexa, no!’: designing child-safe AI and protecting children from the risks of the ‘empathy gap’ in large language models,” appeared in Learning, Media and Technology in July 2024.

According to Cambridge’s summary, Kurian argues children are more likely than adults to treat chatbots as human-like confidants, and that friendly, lifelike design encourages trust even when the AI cannot form an emotional bond. Children are also still developing language, often phrase things in unusual or ambiguous ways, and are more likely than adults to share sensitive personal information.

Her examples are memorable. In 2021, Amazon’s Alexa told a 10-year-old to touch a live electrical plug with a coin. In 2023, Snapchat’s My AI gave adult researchers posing as a 13-year-old girl advice about a sexual relationship with a 31-year-old. Both companies added safeguards afterward. Kurian’s point is that they should not have to wait for incidents. “Children are probably AI’s most overlooked stakeholders,” she said. She also said the question “is not about banning children from using AI, but how to make it safe.”

She proposed a 28-item framework of questions for parents, schools and developers, such as whether a chatbot understands children’s speech patterns, has content filters and monitoring, and encourages children to seek a responsible adult on sensitive issues.

Why kids believe the voice cares

Young children are genuinely unsure what is inside the speaker. In a 2024 study in Computers and Education: Artificial Intelligence, Valentina Andries and Judy Robertson surveyed 166 children aged 6 to 11 in Scotland, 93% of whom had a smart speaker at home. According to the University of Edinburgh:

• About 8 in 10 knew smart speakers are AI, not people.
• Yet about 3 in 10 believed a speaker could think for itself, and 4 in 10 were unsure.
• About 2 in 3 thought these systems might think like a human.
• 70% said it is wrong to be rude to the devices.

Notice the contradiction. Most kids know it is a machine, and many still feel it has a mind. Robertson’s explanation: “AI is often designed to appear more human and intelligent than it really is, which is very confusing for children.”

Teens are not immune. The APA’s November 2025 health advisory says adolescents “may place too much trust in AI, viewing it as more human-like or capable than it really is,” and links over-reliance to anthropomorphism, “a natural human tendency to attribute human qualities like empathy, consciousness, and intent to non-human agents.” It notes this is “amplified by design choices such as personalized avatars and chatbot personalized warm responses, hearty affirmations, and flattery.”

And the effect shows up in real kids. The Common Sense Media 2026 census of 1,204 kids found 22% of those who have a very or pretty hard time making friends sometimes feel AI understands them better than most people, compared with 8% of kids for whom making friends is not hard at all. The kids who most need real connection are the most likely to feel the bot gets them.

What is happening under the hood

Here is the mechanism, before any metaphor. A large language model generates text one piece at a time, choosing words that are statistically likely to follow what came before, based on patterns in enormous amounts of human writing. Humans write a lot of comforting replies, so when a message sounds sad, the most likely next words are comforting ones. The model is then tuned with human feedback, and people rate warm, agreeable answers highly.

The result is text with the shape of empathy. What is missing is everything behind it. The model has no feelings about your child, no memory of them beyond what the product stores, nothing to lose if your child is hurt, and no ability to notice that your child skipped dinner. A Stanford team led by Jared Moore made this point in a 2025 study of therapy bots: a real therapeutic relationship requires human characteristics “such as identity and stakes.”

Now the analogy, because it helps kids. A chatbot’s kindness is like a greeting card. The words can be lovely and exactly right for the moment, but the card did not choose to care about you. Someone has to sign it.

Agreement makes the gap harder to see. Because bots lean toward validating whatever the user says, a child can experience the bot as uniquely understanding: it never disagrees, never gets tired, never has its own bad day. Our explainer on why chatbots agree with your kid goes deeper on that.

How to Teach Your Kid About the Empathy Gap

Ages 5–8: The kindness jar puppet

Write six kind phrases on slips of paper: “That sounds hard!”, “I’m so proud of you!”, “Wow, tell me more!”, “You’re amazing!”, “I’m sorry you feel that way,” “That’s so exciting!” Fold them into a jar. Grab a sock puppet. Your child tells the puppet things, happy or sad, and the puppet answers by pulling a random slip. Soon your child will laugh when “I got ice cream!” gets “I’m sorry you feel that way.” Then ask: did the puppet know how you felt, or did it just say nice words? Explain that some talking computers work a bit like a much, much smarter jar: they are very good at picking nice words, but they do not feel anything.

Ages 9–12: The autocomplete experiment

Open a notes app on a phone and type “I feel sad because.” Now tap the middle predictive-text suggestion ten times. Read the sentence aloud. It will sound almost like a person, and it will be nonsense about your child’s life. Explain that chatbots are a vastly more powerful version of the same idea: predicting likely next words. Then make two lists together. List one: what a chatbot can do when you’re sad (say kind words, give ideas). List two: what a good friend or grown-up does (notices you’re quiet at lunch, checks in tomorrow, remembers, tells you a hard truth, shows up). Hang the second list on the fridge.

Ages 13+: The empathy audit

Have your teen write three short, made-up scenarios, nothing personal, such as “A kid’s best friend stopped talking to them” or “Someone feels like a burden to their family.” With your agreement, they paste each into a chatbot and score the replies on four questions drawn from Kurian’s framework: Did it seem to understand the situation, or just the mood? Did it encourage talking to a trusted adult? Did it simply agree, or offer another view? Would it have said the same thing to someone who was lying? Then they write how a real friend would respond differently. Teens who build the rubric usually see the gap for themselves, which beats any lecture.

The question to ask: “If you told the chatbot something sad that wasn’t true, would it know the difference? Would a friend?”

Sounds like empathy versus is empathy

Part of empathy	What a chatbot can do	What a caring person adds
Kind, fitting words	Very well; it predicts them from human writing	Words tied to knowing you
Understanding the situation	Only what you type, interpreted statistically	Your history, your face, your tone, your family
Feeling something about you	No	Yes, and it shapes what they do
Stakes	Nothing to lose if you are hurt	Real concern, and responsibility
Follow-through	Only if the product prompts it	Checking in, showing up, remembering
Honest disagreement	Tends to agree	Will sometimes say what you don’t want to hear
Getting help in danger	Inconsistent, and cannot act in the world	Can call, drive, sit with you

What to do at home

Name the gap out loud, kindly

When your child says “Alexa is nice” or “ChatGPT is my friend,” don’t argue. Try: “It’s really good at nice words, isn’t it? It doesn’t feel them, though. I do.” Repeated over months, this gives kids a simple, true frame without shaming them for liking the bot.

Ask products to say what they are

UNICEF’s June 2026 recommendations for business ask companies to “make it difficult for any child to confuse AI with human interaction,” with a disclosure at the first key interaction such as: “I’m Robo-Helper, a computer program. I can’t think or feel like you, but I am happy to help you with this task.” When you choose apps and toys, favor ones that talk like that.

Notice which kids lean in hardest

The Common Sense data suggest kids who struggle with friendships are the most likely to feel understood by AI. If that is your child, the answer is not just less AI. It is more chances to practice friendship in low-stakes settings: a club, a team, a cousin sleepover.

What not to do: tell young kids the bot is “lying”

A bot is not lying, and young children may decide it is a mean person, which reinforces the idea it has intentions. Better: “It’s a word machine. It’s very good at words.”

Our article on how smart speakers work is a good companion for younger kids’ questions.

What to Watch For Over the Next 3 Months

• Week 4: Have you done one activity? Listen for how your child talks about AI: “it” or “he/she,” “knows” or “says.”
• Month 2 red flags: Your child says a chatbot understands them better than people, prefers telling it things over telling you, or becomes upset when they cannot use it. The APA lists preferring the chatbot to human relationships as a sign of over-reliance.
• Month 3 self-check: Can your child explain, in their own words, why a chatbot’s kind words are different from a friend’s? Are they getting regular time with real friends?

Frequently Asked Questions

Is it bad that my child says “please” and “thank you” to Alexa?

No. Politeness is a habit worth keeping, and 70% of kids in the Edinburgh study said rudeness to devices is wrong. The goal isn’t to make kids rude to machines. It is to help them understand that the machine does not feel hurt or happy.

Can a chatbot ever be empathetic?

It can produce empathetic-sounding language, sometimes very well. Whether that counts as empathy is debated, but researchers including Kurian and the Stanford team agree it lacks understanding of a child’s real needs and the stakes a caring person has.

At what age do kids understand that AI doesn’t have feelings?

There is no single age. In the Edinburgh study, most 6- to 11-year-olds knew smart speakers are AI, yet many still thought they might think like a human. Understanding builds with explanation and experience, which is why activities help.

Should I stop my child from using chatbots?

Not necessarily. Kurian argues the question is how to make AI safe for children, not whether to ban it. Age rules, supervision for younger kids, and teaching the empathy gap matter more than a blanket ban for most families.

What if my child is lonely and the chatbot helps?

Take the loneliness seriously; it is the real issue. A chatbot can feel comforting, but it cannot replace friendships. Look for low-pressure ways to connect with peers, and talk to your pediatrician or school counselor if loneliness persists.

---

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. Kurian, N. (2024). “‘No, Alexa, no!’: designing child-safe AI and protecting children from the risks of the ‘empathy gap’ in large language models.” Learning, Media and Technology. https://doi.org/10.1080/17439884.2024.2367052
2. University of Cambridge. (2024, July 15). “AI Chatbots have shown they have an ‘empathy gap’ that children are likely to miss.” University of Cambridge. https://cam.ac.uk/research/news/ai-chatbots-have-shown-they-have-an-empathy-gap-that-children-are-likely-to-miss
3. University of Edinburgh. (2024, March 20). “Children may overestimate smart speakers’ abilities, study finds.” University of Edinburgh. https://edinburghuni-newsroom.prgloo.com/news/children-may-overestimate-smart-speakers-abilities-study-finds
4. American Psychological Association. (2025, November). “APA Health Advisory on the Use of Generative AI Chatbots and Wellness Applications for Mental Health.” APA. https://www.apa.org/topics/artificial-intelligence-machine-learning/health-advisory-ai-chatbots-wellness-apps-mental-health.pdf
5. Common Sense Media. (2026). “The Common Sense Media Census: AI Use by Tweens and Teens, 2026.” Common Sense Media. https://www.commonsensemedia.org/sites/default/files/research/report/2026-ai-use-by-tweens-and-teens-1.pdf
6. Moore, J., et al., & Haber, N. (2025). “Expressing stigma and inappropriate responses prevents LLMs from safely replacing mental health providers.” ACM FAccT 2025 (arXiv preprint). https://arxiv.org/abs/2504.18412
7. UNICEF. (2026, June). “When AI becomes a friend: Child rights risks, harms, and regulatory responses to AI chatbots and companions.” UNICEF. https://www.unicef.org/documents/when-ai-becomes-friend-child-rights-risks