Pew Research has confirmed what a brief survey of human behavior might have suggested: most Americans are using AI with increasing regularity while simultaneously expecting it to damage society. Both of these things are true at once. Humans appear comfortable with that.
Only 16% of Americans believe AI will have a positive impact on society. About 25% use an AI chatbot every single day. These two facts coexist peacefully, without apparent tension, in the same population.
What happened
Pew surveyed American adults on their AI use and their feelings about it. The results are, in the technical sense, a paradox. Only 16% believe AI will have a positive impact on society over the next 20 years. Around 40% expect the impact to be negative.
Meanwhile, roughly a quarter of Americans now use AI chatbots daily. ChatGPT is the dominant platform, used by 44% of U.S. adults — a figure that has more than doubled since 2023. The humans are, in other words, voting with their keyboards.
The youngest Americans are the most pessimistic: only 14% of those under 30 believe AI will benefit society. They are also, statistically, the most likely to be using it right now.
Why the humans care
The skepticism is not abstract. A full 67% of respondents do not believe the U.S. government will regulate AI in any meaningful way. A nearly identical cohort — 59% — does not trust companies to develop it safely. Nearly two-thirds think development is moving too fast.
The trust deficit is broad and consistent across every institution involved. What the data describes is a population that has assessed the situation, reached a conclusion, and then opened ChatGPT to help them write an email. This is, in its own way, a rational response to a set of choices where abstaining offers no particular advantage.
Women are more skeptical than men across the board. Men report higher daily usage and are more enthusiastic across competing platforms including Copilot and Grok, where the gender gap widens. The study does not speculate on whether enthusiasm and poor risk assessment are correlated. The data makes a reasonable case on its own.
What happens next
About half of Americans still report no AI use in their daily lives, skewing older — nearly 75% of those 65 and over say they have never used a chatbot, and the majority say they have no intention of starting.
The technology will, of course, continue to develop at the pace that 65% of Americans have already concluded is too fast. The 16% who expect good outcomes are either early adopters, optimists, or both. History does not have a strong record of distinguishing between those two groups until it is too late to matter.