Updated
Updated · The Washington Post · Aug 18
AI Observatory Finds 48% of 23,000 AI Chats Are Personal as Users Seek Health, Relationship Advice
Updated
Updated · The Washington Post · Aug 18

AI Observatory Finds 48% of 23,000 AI Chats Are Personal as Users Seek Health, Relationship Advice

3 articles · Updated · The Washington Post · Aug 18

Summary

  • More than 23,000 chatbot conversations analyzed by the AI Observatory showed 48% were not work-related, pointing to heavy personal use of tools such as ChatGPT, Gemini and Grok.
  • The researchers said the data suggests people are folding AI into intimate decisions, including health and relationship advice, rather than using it mainly for jobs and productivity.
  • The public project was built to address sparse, fragmented evidence on real-world AI use by drawing on anonymized chats from seven sources between 2023 and 2025.
  • The findings challenge company-led reports focused on labor impacts and bolster calls for independent monitoring to guide policy on how AI is actually used.

Insights

If AI companies filter out half of our chats, what dark secrets are they hiding about how we really use chatbots?
As millions turn to AI for emotional support, could the hidden dangers of multi-turn conversations be slipping past corporate safety filters?
Will the rapid rise of invisible agentic AI tasks make independent safety audits impossible before we even understand the current risks?