Pew Research surveyed 5,000 American adults in June 2025 and found a picture that the AI industry would rather not discuss. Only 16% of Americans believe AI will have a positive impact on society over the next 20 years. Around 40% expect it to be harmful. The share who say they are more worried than excited about AI has grown from 37% in 2021 to 50% today.
None of this is stopping people from using it. About a quarter of Americans now use AI tools every day. ChatGPT alone is used by 44% of U.S. adults, more than double the share from 2023. People are logging in, running searches, asking questions at work. Then they close the browser and tell a pollster they are worried about where this is all going.
This kind of split is not unusual with new technology. People drove cars before seatbelts existed and smoked cigarettes while reading about cancer studies. What is notable here is the speed: AI has reached mass use faster than almost any technology in history, yet public sentiment has moved in the wrong direction during that same period. More use, more concern, at the same time.
The trust gap goes beyond individual feelings. Two-thirds of Americans do not believe the government will regulate AI in a meaningful way. More than half do not trust companies to develop it safely. A coalition of 42 state attorneys general sent a letter in late 2025 to major AI companies, including OpenAI, Google, Meta, and Anthropic, raising direct concerns about unreliable and misleading outputs to users. The skepticism in the Pew data has institutional echoes.
Globally, the picture is different. In China, Indonesia, and Thailand, large majorities say AI is more beneficial than harmful. The U.S. and Western Europe are outliers: the U.S. registers AI trust at roughly 32%, compared to around 67% in the UAE. American pessimism is a cultural and political product as much as a rational one, shaped by media coverage, job anxiety, and a general distrust of large institutions.
Younger people, those under 30, are the most negative group. Only 14% of them expect AI to benefit society. This matters more than it might seem. These are the workers entering your organization right now, the customers you are trying to retain, the employees who will decide whether to use the AI tools you roll out. Their resistance is not based on ignorance: they are heavy users. It is based on what they see.
For business operators, the practical read is this. Your customers and staff are likely already skeptical of AI, even if they use it. That changes how you should present AI-assisted services. Positioning something as powered by AI is not automatically reassuring. In some sectors, naming it at all may increase friction rather than reduce it. The smarter approach is to let outcomes speak: faster replies, fewer errors, better service. The label matters less than the result.
The other thing worth tracking: 53% of Americans now say AI will reduce their ability to think creatively. Half believe it will harm their ability to form meaningful relationships. These are not abstract fears about the future. They are present-tense concerns about what the technology is already doing to daily life. A workforce that feels diminished by the tools it is handed will not use those tools well. That is an organizational problem, not just a technology problem.