How US Youth Use AI Chatbots: Conversation Patterns From Naturalistic Keystroke Observations

Editorial reading guide · Kindness

What might my child be seeking when an AI conversation feels supportive or friend-like?

Shows what youth type into contemporary GenAI apps, including social practice, friend-like exchange and emotional support. Tool use dominated app-days; emotional-support codes were relatively uncommon but involved longer text. These observations can start a curious family discussion about the needs behind chatbot use; they do not show that AI provides effective care.

Study context

3363 US users of the paid Aura parental-monitoring app with its optional keyboard enabled; 42,355 user-app-days, January–September 2025.

Age / stage: Parent-reported ages 1–17; most users 10–17; ages below 10 especially uncertain

Read with three questions

1. How do percentages of users differ from percentages of user-app-days when judging how common emotional support is?
2. What can outgoing keystrokes reveal about a child’s needs when the researchers cannot see the chatbot’s replies?
3. How might paid parental monitoring, optional keyboard use and uncertain birth dates change the apparent age patterns?

Keep the limits in view

Only outgoing mobile-app text was observed, with no chatbot responses or psychological outcomes. The selected monitored sample is not nationally representative. Parent-entered birth dates, shared devices and missing keyboard periods complicate age comparisons; toddler entries cannot establish toddler use. LLM coding with human checks may still misclassify themes.