Prevalence of Mental Health Discussions in Publicly Available Generative AI Conversations

Artificial intelligence (AI) chatbots are increasingly used for social, relational, and emotional guidance, as they provide advice in situations ranging from everyday stress to explicit mental health crises.1 Technology companies, researchers, and news outlets have begun estimating how often these interactions occur, but their estimates diverge sharply. One AI developer has concluded that 0.01% of messages in a given week “indicate possible signs of mental health emergencies related to psychosis or mania,”2 while Common Sense Media has reported that 72% of teenagers have used AI chatbots as “companions,” a framing that implies a broad social-emotional role for AI in adolescents’ lives.3 These figures may not represent competing measurements of a single phenomenon; they may reflect different operational definitions of what counts as mental health use. We aimed to quantify how prevalence varies under conservative vs expansive definitions of mental health by using a large public dataset of chatbot transcripts.