The recent report from Imagining the Digital Future Center sketched the lay of the land for large language models (LLMs) like ChatGPT, Copilot, Claude and Gemini. It was particularly notable for mapping the size of and shape of the LLM user population and the way they were employing these artificial intelligence tools. Half of American adults use the models and many more use these artificial intelligence (AI) systems for personal enrichment and fun than use them for work-related purposes.

Understandably, those big and surprising data points got a wave of attention. Still, there are a number of other insights in the survey findings worth surfacing because they have important implications:

The omnipresence and diversity of LLMs and generative artificial intelligence: The models are not just standalone tools with separate interfaces and APIs for many LLM users. They are also built into digital apps they already use. Here are the contours of that: