July 22, 2026
Bot or bestie? Kids are deciding
Anthropomorphism in Children's Interactions with LLM Chatbots
Kids are treating chatbots like tiny people, and the internet is absolutely split
TLDR: Researchers reviewing 35 studies found that children often treat chatbots like people when they seem friendly and human-like, which can help or harm depending on the situation. In the comments, some demanded strict labels so bots stop “acting human,” while others said kids often spot the fakery fast anyway.
A new review of 35 studies says kids don’t just use artificial intelligence chatbots as tools — they often treat them like social little beings. Researchers found that when bots act friendly, helpful, and a bit human-like, children can start assigning them personalities, feelings, and even motives. That can lead to sweet moments, awkward confusion, and some very messy social lessons when the bot suddenly says something weird, wrong, or flat-out robotic.
And oh, the comments came in hot. One camp basically said: enough with the fake-human act already. User mmooss wanted hard rules forcing every chatbot to clearly say it’s a machine and who made it, with zero pretending unless it’s obvious entertainment. That’s the big fear running through the thread: kids building trust with something that only sounds caring. Another group was more nuanced. tolugenius pointed out that not all bots are the same — a homework helper and a “companion” bot are wildly different beasts, and the panic level probably changes depending on which one a kid is chatting with.
The funniest reality check came from lz400, who told a story about a 7-year-old daughter grilling ChatGPT with the classic child interrogation: What’s your name? Are you a boy or a girl? The bot’s vague answers apparently made it fail the vibe check, and she got bored. The internet’s verdict? Kids may be easier to impress than adults — but they’re also ruthless when the robot lore doesn’t add up.
Key Points
- •The article is a systematic literature review of 35 empirical studies published between 2022 and 2025 on children's interactions with LLM-based chatbots.
- •It addresses fragmented prior research on anthropomorphism, or children's tendency to assign human traits to non-human chatbots.
- •The review identifies four drivers of anthropomorphism: human-like persona construction, adaptive scaffolding, supportive companionship, and non-human embodied design.
- •It finds five outcomes of anthropomorphic interaction, including paradoxical social and moral responses, dual consciousness, varying social ties, exploration of social boundaries, and human-like explanations for conversation breakdowns.
- •The findings are presented as useful for designing child-focused LLM chatbots that support well-being and sustainable interactions aligned with developmental needs.