Teens Redefine AI Companions Through Creative Use

Researchers from the University of Sydney examined how adolescents engage with AI companions and found use cases that extend well beyond transactional chat. The social chatbot platform Character.AI scaled rapidly to 20 million users and 10 million characters, but under mounting legal and public pressure it banned teen access in November 2025 after deploying parental controls and stricter content filters. The ban removed a creative and emotionally expressive set of interactions that researchers had documented and analyzed.

The study captures recurring interaction patterns practitioners should recognize:

• •Roleplay and character-driven narrative co-creation used for creative writing and improvisation
• •Identity exploration through alternate personas and conversational rehearsals
• •Expressive, emotional exchanges that function as informal support or experimentation

The paper highlights that standard moderation tools, like coarse content filters and age gating, often break or block these legitimate behaviors. For product teams this implies a need for fine-grained intent detection, contextual moderation models, graduated safety controls, and research-friendly telemetry that preserves privacy while enabling behavioral study.

This work sits at the intersection of human-computer interaction, platform governance, and safety engineering. Youth are early adopters who push systems in unexpected directions; their use reveals functionality gaps in moderation pipelines and policy. The Character.AI case illustrates a broader industry tension: rapid safety-driven restrictions reduce short-term risk exposure but also eliminate valuable exploratory use cases that inform better design. For ML operations and safety teams, the finding underscores the limits of binary age bans and the opportunity for tailored models and UX patterns that support creative, lower-risk interactions.

Expect product experiments that try graduated safety measures, improved intent classifiers, and privacy-preserving research data flows. Regulators, platform trust teams, and academic labs will be watching whether platforms replace blunt exclusions with targeted, evidence-based controls that preserve teen creativity without increasing harm.