Artificial intelligence (AI) is becoming part of everyday life in the U.S., but Americans are far less comfortable when it shifts from something they use to something that makes decisions about them, according to a new survey led by Rutgers researchers.
Nearly three-quarters of U.S. adults (72%) have used at least one major AI tool, up from 50% in November 2024, when researchers conducted an earlier survey of 4,767 people. The findings come from the latest report from the National AI Opinion Monitor, a project led by Rutgers researchers that tracks public attitudes toward AI.
More than a quarter of Americans (27%) use AI every day, yet fewer than one in 10 would allow it to make the final call without human supervision on hiring, loans, college admission, parole or priority for a medical procedure. Between 45% and 54% say AI shouldn’t be used at all in those decisions.
“We adopt AI because it is useful, convenient or expected of us,” said Katherine Ognyanova, a professor with the Rutgers School of Communication and Information and a coauthor of the report. “At the same time, people are concerned about AI taking over high-stakes tasks or making important decisions without human supervision. Overall, using a tool and trusting it fully are two different things.”
For many Americans, AI is becoming part of both their personal and professional lives. Sixty-eight percent use it for personal purposes, up from 45% in late 2024. Among employed adults, 44% use AI for work, up from 32%, and nearly a third (31%) say their job requires it.
The stakes are particularly high when AI is used in decisions involving jobs, finances, education and health. The survey found people were more open to AI assisting with some high-stakes decisions than making them on its own. For example, 34% would allow AI to assist with hiring decisions without making the final call, while 26% would allow it to assist with parole decisions.
“People seem to draw the line at letting AI make unsupervised decisions that significantly affect human lives,” Ognyanova said. “People are somewhat comfortable with a role for AI, just not letting it make the final call.”
Americans also want to know when AI is being used to evaluate them. Seventy-six percent say employers should disclose when AI evaluates workers or job applicants, 76% want companies to disclose when AI makes decisions about customers and 77% want schools to disclose when AI evaluates students. Fewer than 6% disagree with any of those requirements.
Most Americans also favor stricter government oversight of AI. Fifty-nine percent say AI should be strictly regulated because of its potential risks, compared with 25% who favor minimal regulation to promote innovation.
Concern about AI remains widespread. Half of Americans describe themselves as worried about the technology, while the share who say they are angry rose from 19% in late 2024 to 30% in August 2026. Worry isn’t limited to those who avoid AI: Forty-eight percent of daily users say they are worried about it.
People seem to draw the line at letting AI make unsupervised decisions that significantly affect human lives,” Ognyanova said. “People are somewhat comfortable with a role for AI, just not letting it make the final call.
Katherine Ognyanova
Attitudes toward AI also have shifted along party lines. In late 2024, 56% of Democrats expressed confidence in AI to act in the public interest, compared with 45% of Republicans. By August 2026, the numbers had essentially flipped: 44% of Democrats expressed confidence compared with 56% of Republicans. Overall confidence in AI barely changed, remaining at 46%.
“One surprising finding is the partisan reversal on AI,” Ognyanova said. “While overall confidence barely moved, the two groups have switched places. Notably, both parties seem to agree on policy preference: 62% of Democrats and 57% of Republicans still want AI strictly regulated.”
Ognyanova said the survey also revealed a gap between how much people know about AI and how much they think they know.
Fifty-nine percent of Americans say they understand the basics of how AI works, up from 50% in late 2024. But when researchers asked respondents to judge whether 12 statements about AI were accurate or inaccurate, the average score was 48% correct. Among daily users, 82% say they understand the basics, while averaging 49% correct.
“Use appears to increase confidence considerably faster than it increases knowledge,” Ognyanova said.
Researchers surveyed 5,154 U.S. adults between July 5 and Aug. 1, 2026. The surveys were conducted online by market research technology company PureSpectrum using a nonprobability sample with representative quotas for gender, age, race, ethnicity and region. Researchers reweighted the data to match the U.S. population across gender, age, race, ethnicity, education, region, internet use and whether respondents live in urban, suburban or rural areas.
The report was coauthored by Vivek Singh, professor of library and information science at Rutgers; Eunbin Ha, a Rutgers doctoral student; and Ata Uslu, a doctoral researcher at Northeastern University.
The National AI Opinion Monitor has collected three waves of national survey data between October 2024 and August 2026 to track how Americans use and understand AI, where they believe it should and shouldn’t be used, how much they trust the technology and their views on its regulation.
Learn more about the Rutgers School of Communication and Information.
