The better Americans understand artificial intelligence, the less they seem to like it. That’s the picture emerging from new polling by Gallup, which finds that growing familiarity with AI is being matched – and even outpaced – by rising skepticism.
Gallup reports that roughly seven in ten U.S. adults now describe themselves as either somewhat or extremely knowledgeable about AI, an increase from 64% in 2024. What might look like good news for technology adoption is, however, paired with a cooling of public attitudes. Rather than making people more enthusiastic, exposure and awareness appear to be sharpening concerns about the technology’s risks.
A growing share of Americans now say AI is a net negative. Thirty‑nine percent believe artificial intelligence does more harm than good, up from 31% in 2025. Only 9% take the opposite view, saying AI does more good than harm. The plurality, 52%, sits uneasily in the middle, judging that AI brings equal amounts of benefit and damage.
That middle ground is crucial: it suggests that most people have not written AI off entirely, but they see its promise as tightly bound to its potential for misuse. Rather than viewing AI as a purely transformative opportunity or an unambiguous threat, Americans are increasingly treating it as a powerful but double‑edged tool.
The survey also finds that rising skepticism isn’t happening in a vacuum. As Americans learn more about concrete AI applications – from chatbots that can pass for humans to image generators creating realistic deepfakes – their worries are becoming more specific. Concerns cluster around three main themes: the future of work, corporate responsibility, and the broader societal impact of the technology.
On jobs, more Americans now expect AI to reduce the total number of positions available in the U.S. over the next decade. Automation was once seen as a blue‑collar issue, but AI’s ability to handle tasks in customer service, marketing, software development, law, and even medicine is convincing many that white‑collar and creative roles are also at risk. The fear is not just about robots in factories, but algorithms in offices and cloud‑based services quietly replacing human decision‑making.
Trust in business is another pressure point. The poll indicates that people are becoming less willing to assume companies will deploy AI responsibly. High‑profile mishaps – biased algorithms, opaque recommendation systems, and tools that accidentally expose personal data – have made “move fast and break things” an alarming motto in the AI era. For many Americans, it is not the abstract concept of AI that feels dangerous, but the way powerful institutions might use it to cut costs, monitor behavior, or manipulate consumers.
The broader social impact of AI is also coming into sharper focus. Respondents are increasingly exposed to stories about AI‑generated misinformation, impersonation scams, and fraudulent content that is almost impossible to distinguish from reality. This undermines trust in what people see and hear online, feeding a sense that AI could destabilize public discourse and make it harder for citizens to agree on basic facts.
Interestingly, Gallup’s numbers show that increased knowledge does not simply map onto greater acceptance. Instead, familiarity seems to give people a richer mental model of AI’s capabilities and limitations – and that deeper understanding is often sobering. As tools become more visible in everyday life, Americans can better imagine how they might affect their jobs, privacy, and political environment, and they are recalibrating their attitudes accordingly.
The poll hints at generational and educational differences, though the most detailed breakdowns fall beyond the core topline results. Historically, younger adults and more highly educated respondents tend to report higher awareness and more direct exposure to AI tools at work or school. Yet they are not immune to the broader trend: even among more “tech‑savvy” groups, rising knowledge is increasingly accompanied by sharper criticism around fairness, transparency, and long‑term risks.
For policymakers, these findings are a warning sign. While governments and companies race to integrate AI into healthcare, education, security, and public services, public opinion is moving toward caution. If people come to believe the harms are outpacing the benefits – or that there are no effective safeguards – resistance to AI‑driven reforms could harden, from workplace rules to data‑sharing initiatives.
Businesses face a similar challenge. Many executives are betting that AI will unlock cost savings and productivity gains, but Gallup’s data suggests a trust deficit is forming. Employees may interpret AI deployments primarily as a prelude to layoffs or increased surveillance. Customers may worry that algorithms will be used to upsell, nudge, or profile them in ways they cannot see or contest. Without meaningful transparency and clear guardrails, the commercial upside of AI could be undermined by client and worker pushback.
AI developers and startups should read the poll as a design and communication mandate. The public is telling them that “cool demos” are no longer enough. People want to know: Who is accountable when the system gets it wrong? How is personal data protected? What biases have been tested for, and how are they mitigated? Tools that are explainable, auditable, and clear about their limits will likely fare better in a climate of growing skepticism.
At the same time, the numbers leave room for a more constructive narrative. With 52% of Americans still seeing AI as equally good and bad, and only a small fraction firmly convinced it is all harm, there is a substantial “persuadable” majority. Concrete examples of AI being used to detect fraud, assist doctors, improve accessibility, or support scientific research could help rebalance the conversation – but only if they are accompanied by credible protections against abuse.
Education will play a central role in shaping where attitudes go next. As AI literacy programs expand in schools, universities, and workplaces, the quality of that education matters. If it blends technical basics with honest discussion of ethics, safety, and rights, it can create more informed, critical users who are capable of both leveraging AI and demanding better from it. If, instead, it functions purely as a marketing pipeline for new tools, public suspicions are likely to deepen.
Regulation is another key lever. Clear rules around data use, transparency, accountability, and redress can reassure citizens that AI will not be left to self‑police. The more people believe there are enforceable standards – and that regulators understand the technology – the more they may be willing to give AI a chance in sensitive domains like finance, healthcare, and public administration.
Importantly, the Gallup findings underline that “public opinion on AI” is not a static snapshot but a moving target. Over the past few years, sentiment has swung from curiosity to optimism, and now toward increased wariness as real‑world deployments have exposed both promise and peril. Future shifts will likely follow the same pattern: spectacular breakthroughs or visible benefits could soften views, while scandals, abuses, or economic shocks connected to automation could harden them.
In the end, Gallup’s new numbers capture a pivotal moment. Americans now know more about artificial intelligence than ever, but that knowledge is coming with strings attached. The technology’s future acceptance will depend less on technical advances alone and more on whether companies, developers, and lawmakers can align AI’s deployment with the public’s expectations of fairness, accountability, and shared benefit.