United Kingdom 
United States 
Deutschland 
France 
España 
Italia 
International 
Two special edition episodes of Reality checks capture America's mood toward and adoption of artificial intelligence in 2026.
Recorded live at HumanX 2026, a conference bringing together 6,000 of the most influential leaders in AI, the new Reality checks episodes feature:
- Gina King, investor and partner at Supernode Global
- Taylor Lorenz, technology and online culture journalist, and founder of User Magazine
To prepare for their interviews, both guests received copies of a five-question survey sent to 1,000 Americans. YouGov conducted the survey online on Tuesday, April 7. Gina King and Taylor Lorenz predicted the results of the survey in the interviews below the very next day, Wednesday, April 8.
The survey
- How often do you use any form of artificial intelligence (AI) tools (e.g., chatbots, recommendation systems, voice assistants, agents, etc.)?
- Which of the following reasons have you used conversation-based AI (e.g. ChatGPT, Claude, Google Gemini, etc)? Please select all that apply.
- If you were using an AI agent to shop for you, how much would you be comfortable with it spending without your final approval?
- Generative AI tools like ChatGPT or Claude are built on LLMs. What does LLM stand for? Please answer without looking it up.
- Please rank the following based on how “AI-like” they feel to you, from most “AI-like” (1) to least “AI-like” (5).
More than four in five Americans say they use AI every day
Given examples of generative AI and voice assistants, most Americans say they’ve used artificial intelligence at some point (81%).
When guessing the results in her interview, Lorenz said, "My first inclination was to say like, maybe 60% have never used, but then I consider the fact that ChatGPT was the fastest growing consumer internet product of all time, and they have hundreds of millions of users."
Nearly half of Americans said they use AI technology weekly (48%).
"The conversational AI part of our lives has already permeated," said King. "People are used to interacting with devices and with bots."
Despite this widespread usage, habitual adoption has not yet caught on with the masses, as only 18% of Americans report using AI daily. Millennials stand out as the most frequent users, with one in four saying they use AI every day (25%).
A broader generational pattern is evident: adoption declines with age, with Baby Boomers the least likely to use AI regularly and also the least likely to report using the technology.
"I thought the Baby Boomers were more online, but it's probably good that they're not" said Lorenz.
Yet only a third of Americans know what LLM stands for
As generative AI tools become mainstream, just 34% of Americans correctly identified “LLM” as “Large Language Model,” while nearly half say they’re not sure (48%).
"You don't need to understand what LLM necessarily stands for to use an LLM or understand the purpose of it," said Lorenz.
This suggests a knowledge gap between usage and literacy, where Americans engage with generative AI tools like ChatGPT or Claude without a clear grasp of the underlying technology.
Gen Z is most likely to know what LLM stands for
Even among frequent users, understanding of AI terminology is far from universal. 41% of daily AI users correctly identified “LLM” in our April 7, 2026, survey, compared with 34% of the general population.
LLM awareness notably drops sharply with age. Gen Z leads at 60% correct, compared to 18% of Baby Boomers. With 44% correct, Millennials also sit above the 34% national average.
"The average consumer focuses more on the experience and 'Does it work?' or "Does it not work?" and less about mechanics," said King. "It's always been that way with tech."
Americans are most likely to have used AI to summarize long content
There’s no single dominant use case for conversational AI, as many users say they use it to summarize content (25%) as say none of the somewhat arbitrary listed reasons apply.
Everyday, practical tasks in the survey lead the way, with recipe-finding emerging as a notable use case (19% overall), particularly among women (23% vs. 14% of male users).
More sensitive applications, like medical advice (16%) or life decisions (15%), are not far behind.
"I got a medical diagnosis actually that the AI discovered," said Lorenz. "I went to two specialists to confirm it, because I was very paranoid that it was giving me some sort of misinformation... I think that it's good to use it and take it to a doctor."
Comparing responses from men and women who use AI, female users are more likely to say they have turned to it for personal or social scenarios, while men are much more likely to say they used it to settle debates.
Overall, the results of this survey suggest that conversational AI appears less defined by one core function and more by a wide mix of specific, situational needs.
Despite widespread AI usage, agentic shopping faces skepticism
Consumers remain wary of handing over spending control to autonomous agents, even if a subset trusts AI with sensitive information like medical diagnoses.
Most Americans say they would not let an AI shop for them at all (56%), while another 24% insist on approving every transaction, leaving relatively few comfortable with any level of independent spending.
"If the agent is recommending a product, or you instructed the agent to find a product, put it in your shopping cart, and you have to go and hit pay, there's really low risk to that," said King.
Even among those open to agentic spend, most cap that trust at small amounts, with just 10% allowing purchases higher than $25 without approval.
While Millennials are somewhat more open than other generations, the broader pattern suggests agentic shopping adoption lags well behind general AI usage, with trust - especially financial - still a major barrier as the functionality enters the market.
Recipe searchers are twice as likely to authorize agentic shopping
While most Americans remain hesitant to trust AI agents with autonomous spending, that resistance softens considerably among those already using AI for recipe discovery.
Just 34% of AI recipe users say they wouldn’t allow agentic shopping at all, compared with 56% of the general population, and they are notably more comfortable with higher spend thresholds (22% willing to allow $25+ without approval vs. 10% of the U.S. general population).
" I would feel comfortable letting it order, like household items, groceries, things like that," said Lorenz. "For the big ticket purchases, I do want stay as a human. It's not going to buy a car for me."
The takeaway for agentic developers and marketers is that familiarity with practical, low-risk AI use cases - like cooking - may help build confidence in more advanced, higher-stakes applications.
Ants and bees are most like AI; Llamas are least like AI
The HumanX interviews ended with a playful question, gauging how Americans relate animals and insects to artificiall intelligence.
While respondents don’t fully agree on what feels “AI-like,” collective, swarm-based creatures seemed most likely, which Lorenz and King both predicted correctly.
Ants are most often ranked as the single most “AI-like” (25% place them first), whiles bees achieve a stronger overall showing, landing higher on average across rankings.
"With AI, you need a lot of data. You need large data sets to train these models," said King. "There's a lot of collectives working together to generate the outputs that we're all seeeking."
Meanwhile, creatures like llamas fall firmly on the opposite end of the spectrum, with half of respondents ranking them as least “AI-like.”
The results point to a public intuition that AI is defined more by networks and coordination than by any singular “smart” entity.
"I think of an LLM as kind of a hivemind for humanity," said Lorenz. "It's taking in massive amounts of information and it's a collective knowledge we have access to. We're the ants, maybe."