Only 16% of Americans See AI as Good for Society

Only 16% of Americans believe AI will have a positive impact on society over the next 20 years, according to a new Pew Research survey. About 40% expect the impact to be negative, and the rest land somewhere in between. The gap between Wall Street's enthusiasm for AI and the public's mood has rarely looked this wide.

The skepticism runs deeper than a passing mood. Nearly two-thirds of respondents think AI is being developed too fast. 59% don't trust companies to build it safely, and 67% doubt the U.S. government will regulate it in any meaningful way. Even among people under 30 — the group most likely to actually use these tools — just 14% see a positive future. There is no age bracket where optimists are the majority.

1People use AI every day and still don't trust it

Here's the twist that makes the numbers worth reading: distrust and usage are climbing together. About a quarter of Americans now use an AI chatbot daily. 44% have used ChatGPT — more than double its 2023 share — with Gemini at 24%, Copilot at 17%, and Meta AI at 14%. Six in ten say they routinely read the AI-generated summaries at the top of search results, often without choosing to. At the same time, roughly half never touch AI in daily life, and nearly three-quarters of people over 65 have never used a chatbot at all. Men outpace women on daily use, 27% to 20%.

The headline isn't 'people hate AI.' It's that adoption is being driven by convenience and default settings, not by confidence. Folks reach for ChatGPT because it's faster, then quietly assume it might be wrong, leaking their data, or making things up. That's a fragile foundation. It's exactly the environment where a single high-profile failure — a hallucinated legal brief, a botched medical answer, a scam built on a cloned voice — can sour millions of people in a week. Trust here is thin, and thin trust breaks loudly.

For anyone building with these tools, the survey is a free product brief. The market isn't short on capability — it's short on trust. The features that will win aren't flashier demos; they're the boring ones that show their work: citations, an easy way to check a claim, clear limits on what the thing can and can't do. People are already using AI. They just want a reason to believe it.