Despite regular reports from tech giants like Anthropic and OpenAI, a new study reveals that the real uses of artificial intelligence (AI) are far more diverse than what these companies showcase. According to researchers at Stanford’s Trustworthy AI Research Lab, independent sources are essential for understanding how people truly interact with AI.

Their project, the AI Observatory, gathers and analyzes anonymized conversations between users and popular models like Claude and Gemini. The findings highlight significant variations in AI usage across different platforms. For instance, a greater percentage of conversations revolve around personal topics such as health, relationships, and even sensitive material, rather than work-related tasks.

The study also notes that the nature of interactions has evolved over time, with more users engaging in small talk and less self-disclosure from AI companions. This shift suggests a growing comfort level among users, though it remains to be seen how such conversations are being moderated by platforms.

Differences between models are stark: some, like Grok, see heavy use for information retrieval but also misinformation, while others excel in social and roleplay scenarios. These insights underscore the need for a more comprehensive understanding of AI's impact on society, rather than relying solely on corporate data.