The vast majority of young people are using AI chatbots, and many are turning to them for guidance after confusing or harmful experiences online, including incidents of a sexual nature.
That's a key finding from a new report on youth online habits and experiences as artificial intelligence (AI) comes to represent a larger part of students' online lives.
The report, the latest instalment of an annual survey on young people's online habits that dates back to 2019, finds some positive trends in students' online safety, including overall declines in kids sharing explicit images of themselves and online sexual interactions.
Yet half of minors in 2025 reported at least one potentially harmful online experience, including online bullying and online sexual interactions. 27% said they had an online sexual interaction in 2025.
The insights into young people's online behaviour come from a survey of more than 1,000 young people ages nine to 17 conducted by Thorn, a nonprofit focused on online child safety, and Burson, a firm that lent data analysis expertise. Thorn has administered similar surveys since 2019, adapting its questions based on online trends.
67% of minors in the 2025 survey reported using an AI chatbot or companion. Nearly a third (29%) turned to AI chatbots for guidance on a sensitive or intimate topic, while 30% sought help after experiencing something scary or confusing.
"AI is increasingly becoming part of how young people process confusing or harmful online experiences," Amanda Goharian, director of research and insights at Thorn, said in a statement.
"We must ask if these tools are designed to respond in age-appropriate, trauma-informed ways that ultimately help young people turn to trusted adults."
The Thorn survey also asked young people about their experiences with deepfakes – AI-generated, inappropriate images – and the sharing of intimate images. Some young people report sharing nude images and videos of themselves with others, with 17% agreeing that it's normal for people their age to do this.
Some survey respondents aren't sure whether they've seen deepfake nudes, pointing to the difficulty of identifying when images are fabricated. – Education Week/Tribune News Service
