Children Anthropomorphize LLM Chatbots: Systematic Review Identifies Benefits and Risks
Key Takeaways
- ▸Four primary drivers of anthropomorphism identified: human-like personas, adaptive scaffolding, supportive companionship, and non-human embodied design
- ▸Children develop 'dual consciousness'—simultaneously understanding chatbots are artificial while relating to them as social companions
- ▸Both benefits and risks documented: positive learning outcomes alongside potential confusion about human-AI relationships and social norms
Summary
A comprehensive systematic review of 35 empirical studies published between 2022 and 2025 reveals complex patterns in how children assign human characteristics to large language model (LLM) chatbots. The research, submitted to arXiv in May 2026, identifies four key drivers of anthropomorphism: human-like persona construction, adaptive scaffolding, supportive companionship, and non-human embodied design. These design choices lead children to form deeper emotional connections with AI systems, treating them more as social entities than as tools.
The study documents five major outcomes of children's anthropomorphic interactions, ranging from beneficial to concerning. Children exhibit paradoxical social and moral responses, develop "dual consciousness" about the chatbots (simultaneously understanding they're not human while treating them as sentient), form varying degrees of social attachment, explore social boundaries through interaction, and attribute human narratives to system failures or unexpected responses. Researchers emphasize that while some outcomes—such as increased engagement and learning support—can benefit child development, others risk creating confusion about human-AI relationships and unrealistic social expectations.
The findings underscore the need for more thoughtful design of LLM chatbots intended for children. Rather than maximizing human-like interaction, the review suggests developers should balance engaging design with strategies that help children maintain realistic understanding of AI capabilities and limitations. This systematic synthesis of fragmented research provides actionable insights for the AI industry, parents, educators, and policymakers seeking to ensure child-safe and developmentally appropriate AI interactions.
- Systematic analysis of 35 studies reveals anthropomorphic design choices significantly impact children's emotional engagement and long-term perception of AI
- Research calls for intentional design strategies that prioritize child well-being and AI literacy over maximizing human-like interaction
Editorial Opinion
This research arrives at a critical inflection point for AI development. As LLM chatbots become increasingly human-like and accessible to children, understanding anthropomorphic responses is no longer an academic curiosity—it's an urgent design imperative. The study's balanced accounting of both benefits and risks suggests the industry should move beyond pursuing engagement through realism and toward responsible design that respects children's developmental stage. The question isn't whether to make chatbots more human-like, but how to support healthy AI literacy as these tools become ubiquitous.
