Why Americans Are More Concerned Than Excited About AI

AI adoption has accelerated, but public confidence has not followed. Pew’s latest data shows that 52% of Americans are now more concerned than excited about the increased use of AI in daily life, compared with 37% in 2021. Only 9% feel more excited than concerned. 7

That shift suggests a simple explanation: many people experience AI not as a broadly shared improvement, but as a rapid and opaque transfer of power. Employers, platforms, and technology companies gain new capabilities, while individuals are asked to absorb uncertainty around work, education, privacy, creativity, and control.

The benefits feel abstract; the risks feel personal

AI can help with tasks such as writing, search, coding, and customer support. But those benefits are not equally visible or valuable to everyone. A feature that saves a company money may feel to an employee like a threat to their role. An automated service that is efficient for a business may feel to a customer like a frustrating barrier to a human representative.

This creates an imbalance in perception. The promised upside is often expressed through long-term forecasts about productivity, healthcare, or scientific discovery. The downside appears in everyday experiences: an incorrect answer, a scam, a synthetic piece of content, a workplace monitoring system, or an AI feature added without consent.

Work and education make the anxiety concrete

People do not need to believe that every job will disappear to worry about AI. They may instead fear fewer entry-level opportunities, weaker bargaining power, more workplace surveillance, or roles being reduced to supervising automated systems.

Pew’s latest findings show how widespread that concern is: 71% of Americans expect AI to reduce the number of U.S. jobs over the next 20 years. 5 For younger adults, the issue is especially immediate because education and early career paths are precisely where automation could change how people gain experience.

Generative AI also complicates learning. It can help students and teachers, but it can also shortcut assignments, make assessment harder, and blur the difference between assisted work and independent understanding. When people cannot tell what effort is genuinely human, confidence in the surrounding system declines.

Creativity and ownership have become trust issues

For writers, artists, musicians, and other creators, the argument over AI training is not merely technical. It concerns permission, attribution, compensation, and who benefits when human-made work is used to build commercial systems.

Even useful AI output can therefore feel extractive when creators believe their work was absorbed without meaningful consent. The concern extends beyond ownership: Pew has found that Americans are more likely to expect AI to erode than improve people’s creative thinking and ability to form meaningful relationships. 12

That helps explain the appeal of physical media, paper notebooks, film cameras, in-person events, and other offline activities. These are not necessarily signs that people reject all technology. They can represent a desire for provenance, privacy, intentionality, productive friction, and confidence that a human made the thing being valued.

The available evidence does not establish that AI backlash alone is driving a broad return to retro technology. But the cultural response is understandable: when digital systems feel increasingly automated and opaque, human-made experiences become more distinctive.

Default AI features make people feel they are losing choice

Many users might welcome an AI tool when it is optional, transparent, and easy to control. The reaction changes when AI is inserted into a familiar product by default, routes customers away from human support, or raises questions about how personal data is used.

The distinction is important for product design. Adoption is not the same as acceptance. People may use a system because they have no practical alternative while still resenting the company that imposed it. That resentment can affect retention, brand trust, and willingness to pay.

Distrust extends to the people building AI

Public skepticism is directed at institutions and executives as much as at models. A CNBC/Generation Lab survey of 1,088 Americans aged 18 to 34 found that a majority distrusted each of nine prominent AI industry leaders to act responsibly on AI; 45% expected AI to hurt their careers, compared with 10% who expected it to help.

Dario Amodei has described this as a “crisis of trust,” arguing that people already distrust companies, governments, and the technology industry and suspect that new systems will be used against them. 17

That diagnosis explains why public-relations campaigns are unlikely to be enough. When the underlying question is “Who controls this, and who benefits?”, optimistic messaging can sound like evasion unless people can verify the answer in their own lives.

Infrastructure turns an abstract debate into a local one

The AI boom also makes its physical requirements harder to ignore. Data centers connect AI to electricity, water, land, grid investment, and public policy. Communities may reasonably question whether local costs are certain while promised jobs, lower bills, or other benefits remain uncertain.

This is not automatically opposition to technology. It is a demand for accountability: clear disclosure of impacts, fair contributions to infrastructure, reduced local harm, and enforceable community benefits. If those conditions are absent, resistance becomes a rational response to an uneven distribution of costs and gains.

Why falling trust is a business problem

For AI companies, distrust can limit adoption, retention, willingness to pay, recruiting, regulatory permission, and the ability to build infrastructure. A technically impressive model does not create durable consumer value if customers associate it with layoffs, unwanted features, creative extraction, or unreliable service.

Brian Chesky has argued that Silicon Valley needs more AI products that solve ordinary people’s everyday problems, rather than focusing primarily on enterprise buyers or demonstrations of model capability. That is a product argument as much as a communications argument: people are more likely to support technology when they can clearly identify what it improves in their own lives.

Amodei’s criticism is similarly demanding. He has acknowledged that AI companies have not yet delivered on their biggest promises and argued that dramatic claims about curing cancer are no substitute for actual breakthroughs. The implication is clear: credibility will come from results that are reliable, visible, and broadly shared—not from increasingly ambitious forecasts.

What could improve public sentiment

AI companies that want greater acceptance should focus on five practical commitments:

• Build for real needs. Products should solve recurring problems for ordinary users, not merely showcase technical capability.
• Give people control. Clear opt-in choices, easy opt-outs, transparent data practices, and human alternatives make participation feel voluntary.
• Share the gains. Productivity improvements should translate into better services, lower costs, training, and stronger work—not only higher margins.
• Respect human contribution. Creators need meaningful systems for consent, provenance, attribution, compensation, and rights enforcement.
• Prove the promises. Measurable improvements in health, affordability, education, and living standards will do more for trust than another marketing campaign.

The public is not necessarily asking AI to stop progressing. It is asking who the progress is for, who gets to decide how it is used, and whether the people bearing the risks will share in the benefits. Until companies answer those questions through product design and observable outcomes, faster AI adoption may continue to produce less—not more—public enthusiasm.