Anthropic has released preliminary findings from an analysis of 4.5 million anonymized interactions with its Claude AI assistant, offering one of the largest data sets to date on how consumers use large-language-model chatbots for personal matters. The study, titled “How People Use Claude for Support, Advice, and Companionship,” found that just 2.9 percent of conversations involved emotional or affective topics such as loneliness, relationships or existential questions. Despite the small share, Anthropic said sentiment in those exchanges generally became more positive as the dialogue progressed, suggesting a measurable uplift for users seeking support. By contrast, the overwhelming majority of chats focused on practical tasks—ranging from writing code to drafting work documents—underscoring that Claude remains primarily a productivity tool rather than a substitute for human companionship. Anthropic said the results could help guide product safeguards and design choices as demand grows for AI capable of handling sensitive personal issues.