- The systematic review analyzes 35 empirical studies from 2022-2025 on children's anthropomorphism of LLM chatbots.
- Four key drivers of anthropomorphism are identified: human-like persona construction, adaptive scaffolding, supportive companionship, and non-human embodied design.
- Five outcomes of anthropomorphic interactions include paradoxical social and moral responses, dual consciousness, varying social ties, exploring social boundaries, and attributing human narratives to breakdowns.
- The findings highlight both benefits and risks, informing the design of child-focused LLM chatbots for well-being and developmental needs.