Youth frequently use the internet to seek support from their friends but don’t always get helpful responses. Recent advances in AI technology may be able to help.

Young people find support from AI-generated responses on topics ranging from relationships to physical health. When it comes to sensitive topics, such as suicidal thoughts, they prefer human responses, according to a study from the University of Michigan and Drexel University presented recently at the Conference on Human Factors in Computing Systems in Hawaii.

For decades, social media has increasingly become young people’s main method of peer support exchange. Researchers sought to understand when and how AI can facilitate and assist in such exchanges.

U-M and Drexel researchers analyzed survey responses from 622 people aged 18 to 24 who have experience seeking and giving support online. In the study, participants evaluated blinded human- and AI-generated responses to help-seeking messages (meaning they did not know who wrote the responses). The human responses reflected various types of human support young people may encounter, including peers, adult mentors and therapists.

The findings showed that participants responded positively to the AI-generated responses about relationships, self-expression and physical health, noting that they seemed caring and offered actionable suggestions.

However, they strongly preferred the human-generated response to help-seeking messages about a sensitive topic like suicidal thoughts.

“This type of research is important to understand how AI can help young people in a safe and effective way,” said Jordyn Young, doctoral student at the U-M School of Information and the study’s lead author. “It’s not just what the model says, but also how. Being honest about the limitations of AI’s help-giving capabilities is good, but we need to phrase it carefully—especially in response to sensitive disclosures.”

Since young people access the internet daily, having different ways to share their thoughts—especially with AI-generated responses—could be beneficial for the long term, according to the researchers.

“Our research suggests that AI-powered tools may help youth peers support each other during challenging times by helping them figure out what to say,” Young said.

The study’s authors at Drexel included Laala Jawara, Diep Nguyen, Brian Daly, Jina Huh-Yoo and Afsaneh Razi.