While most people in the U.S. who have used large language models (like ChatGPT) for informal learning, entertainment, and getting information about products and services, 39% of U.S. adults have also tapped into LLMs to source information about physical or mental health.

This insight is brought to us in the brilliantly titled report, Close encounters of the AI kind, from the Imagining the Digital Future Center at Elon University. The principle author of the survey report is the Center’s Director, Lee Rainie, whose name many of you will know from his two+ decade career at the Pew Research Center (and the Pew Internet & American Life Project).

 

 

 

 

 

 

 

 

The most commonly-used LLMs were ChatGPT by far the #1 “brand” of consumer-facing AI tools, followed by Google’s Gemini (with 50% of U.S. adults have used), Microsoft’s Copilot (39%), Meta’s LLaMa (20%), 12% using xAI’s Grok, and 9%, Anthropic’s Claude.

One-half of LLM users said they think the models they use are “smarter” than they are, and 76% of users are satisfied with the LLMs’ performance.

Among users’ positive experiences with LLMs are that 54% of folks said using LLMs has improved their productivity, and 42% said their use improved creativity.

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

The Center’s survey polled 500 users of large language models (LLMs) in late January 2025, using the screening question,

“Do you ever use artificial intelligence (AI) large language models (also known as LLMs or generative AI), such as ChatGPT, Google’s Gemini, Anthropic’s Claude, Microsoft’s Copilot, or an open source LLM?”

The demographics of those using LLMs are shown in the second bar chart, beginning with the top line that 1 in 2 U.S. adults use AI large language models — the same percentage of men and women.

Some important insights here are that,

  • Non-White people (Black and Hispanic) are more likely to have ever used LLMs versus White adults (which may be a positive sign for healing digital divides that prevent many health citizens’ access to health information and services)
  • Younger people tend to use LLMs more than older people, with use declining by aging cohort: the greatest use was found among 77% of people ages 18-29, followed by 71^ of adults 30-49, 42% of those 50-64, and 28% of those 65 and older
  • By income, more wealthier people earning $100,000 or more tend to have used LLMs although the percentage differences between the wealthiest cohort versus those earning under $50K a year was relatively small: 57% compared with 53%
  • More people with college degrees have used LLMs compared with those with high school or less education attainment, and,
  • There are slightly more Democrats and Independents than Republicans who have used LLMs as of January 2025 (54% and 55% versus 49%, respectively.

 

 

 

 

 

Health Populi’s Hot Points:  Here in Health Populi, I’ve long discussed “impatient patients,” seeking convenience, access, control, and even delight in health care experiences, observed in the essay, How AI Could Reshape Health Care — Rise in Direct-to-Consumer Models by Dr. Kenneth Mandl. In this JAMA Viewpoint, Dr. Mandl tells us that,

“A time may be approaching when wise patients demand artificial intelligence (AI)-informed care as a standard of quality,” with AI possibly putting the health system at a crossroads.

Why? Because traditional health care organizations — hospitals, physicians, health plans — being, in Dr. Mandl’s words, “often constrained in innovation by entrenched operational structures.”

Dr. Mandl then explains that technology-driven direct-to-consumer companies, such as Amazon, may be able to fast-track DTC AI-embedded health care. This can be done channeling AI-assisted care through licensed health care professionals “or” autonomous AI agents, he writes.

A key differentiator for the new-tech-DTC/AI purveyors is that “Bit Tech maintains control over their own data ecosystems,” and do not, in Dr. M’s estimation, “cede(d) control of their data assets to EHR vendors.”

Driving engaged health consumers, such as the 39% identified in the Center’s report above, are two key forces:

  1. Declining access to primary care, and,
  2. The shift to consumer-controlled health data.

We can be hopeful and bullish cheering health consumers’ engagement with AI for health care — the new “Paging Dr. Google” research flow — while being mindful and attentive to the potential pitfalls in DTC/health-AI, noted by Dr. Mandl in the graphic I summarized from his essay. These are aligning incentives, sustainability, privacy and security risk management, oversight, and mitigating a digital divide that can exist ex ante (as we gain adoption) or ex post (further exacerbating the digital divide or lack of “techquity.”

This is a major workflow and research area for me in 2025 and ongoing, so do stay tuned to Health Populi as I learn more hands-on and through collaboration on our journey into DTC health baked with AI.