Systematic review finds children form complex, sometimes conflicted bonds with LLM chatbots

A systematic review of 35 empirical studies conducted between 2022 and 2025 finds that children routinely attribute human-like qualities to LLM-powered chatbots, and that this anthropomorphism shapes both how kids interact with the technology and how they process its failures.

What's new

The review, "Anthropomorphism in Children's Interactions with LLM Chatbots: A Systematic Review of Drivers and Outcomes" by Hansinie Madushika Jayathilake and Renkai Ma, synthesizes findings across three dozen studies to identify four consistent drivers of anthropomorphism in children's chatbot use: human-like persona construction (names, voices, personalities designed into the product), adaptive scaffolding (the chatbot adjusting to a child's responses in ways that feel attentive), supportive companionship (the chatbot positioned as a friend or confidant rather than a tool), and non-human embodied design (physical or visual forms — toys, avatars, speakers — that still invite social attribution despite not being humanoid).

On the outcomes side, the authors identify five recurring patterns. As the paper states, "Five anthropomorphic outcomes emerged, including children exhibiting paradoxical social and moral responses, dual consciousness about the chatbots, forming varying social ties, exploring social boundaries, and attributing human narratives to conversation breakdowns." That last point is notable on its own: when a chatbot glitches, breaks character, or gives a nonsensical answer, children in the reviewed studies tended to explain the failure using human narratives — the chatbot was "confused," "tired," or "being difficult" — rather than recognizing it as a technical malfunction.

"Dual consciousness" describes children who can simultaneously know, when asked directly, that a chatbot isn't alive or sentient, while still treating it as a social actor deserving of politeness, loyalty, or emotional investment in the moment-to-moment interaction. The review frames this as a genuinely paradoxical state rather than simple confusion.

Context

The review lands amid a broader wave of scrutiny over AI chatbots and children this year, including lawsuits and regulatory actions tied to chatbot interactions with minors and vulnerable users. Where much of that coverage has focused on acute safety failures, this review takes a different angle: it treats anthropomorphism itself — the tendency to relate to a chatbot as a social being — as the underlying mechanism worth understanding, independent of any single incident. The authors compiled evidence across nearly four years of HCI, developmental psychology, and human-computer interaction research, spanning the period in which LLM-based chatbots moved from novelty to near-ubiquity in children's software (voice assistants, companion apps, educational tools, toys).

Why it matters

Design choices that make chatbots feel more human — warmer personas, more responsive scaffolding, companion framing — are already standard practice across consumer AI products, largely because they drive engagement. This review's contribution is to lay out, systematically, what those choices do to children specifically: they don't just increase engagement, they measurably change how kids interpret the system's behavior, including its failures.

That has direct design and policy implications. A chatbot that fails silently or gives a wrong answer isn't just delivering bad output to a child user — per this review, the child may narrate that failure in human terms, potentially masking the fact that something actually broke. The authors' recommendation — that these findings inform future chatbot design decisions aimed at protecting children's well-being — arrives as regulators and platform builders are already under pressure to define clearer guardrails for AI systems used by minors. A concrete evidence base on how anthropomorphism actually functions in children's cognition gives that conversation something more specific to work from than general concern.