LLM Themes Are Not Observations

# LLM Themes Are Not Observations When researchers use AI to automatically pull out themes or patterns from text data, they're creating something new rather than discovering what's actually there—and this matters because decisions based on those AI-generated categories can be misleading. Think of it like asking an AI to summarize customer complaints: the themes it highlights might say more about how the AI works than about what customers actually care about. If you're using AI-generated categories in your analysis or research, you need to verify they match reality before trusting them to guide decisions.