Americans Use AI Daily But Still Hate It, Poll Finds

The NBC News AI poll is out, and the numbers tell a story that should make every tech optimist uncomfortable: more than half of Americans are now using AI tools regularly, yet the technology ranks below ICE, below both political parties, and just barely above Iran on the national favorability index. Usage is climbing. Trust is not.

More Users, Worse Ratings — The Numbers Don't Add Up

The NBC News/Hart Research Associates poll, conducted with Public Opinion Strategies from February 27 through March 3, surveyed 1,000 registered voters. The findings are striking in their contradictions. Some 56% of respondents said they'd used an AI platform — ChatGPT, Microsoft Copilot, or Google Gemini — in the past two to three months. That's up from 53% in August 2025 and 48% in December 2024. A steady climb.

But ask those same people how they feel about AI? Only 26% say they view it positively. A full 46% hold a negative view, putting AI's net favorability at minus 20 points — a figure that trails ICE at -18, President Trump at -12, Kamala Harris at -17, and even California Gov. Gavin Newsom at -18. The Democratic Party lands at -22. Iran clocks in at -53. AI sits comfortably in that bracket.

Call it what you want — a trust deficit, a perception problem, a PR crisis — but the gap between behavior and belief is the actual headline here. Americans are using these tools anyway. That's not enthusiasm. That's dependency.

Why Are So Many People Still Skeptical?

What do Americans actually think about AI risks vs. benefits?

Fifty-seven percent of poll respondents said AI's risks outweigh its benefits. Only 34% said the opposite. Those numbers become sharper when you factor in findings from the Pew Research AI survey published last September — 50% of U.S. adults said they were more concerned than excited about AI, up from 37% just four years earlier. Concern is trending upward faster than usage.

The fear isn't abstract. A Quinnipiac University poll from April 2025 found that just 4% of Americans believe they can trust AI-generated information almost all the time. Nearly three-quarters said the government should step in to prevent AI-caused job losses. That's not skepticism of a new gadget — that's a call for regulation from people who feel the economic ground shifting beneath them.

The YouGov AI trust poll from December 2025 put it bluntly: 35% of Americans use AI at least once a week, but only 5% say they deeply trust it. Trust is lowest in healthcare and finance — which happen to be the two sectors where AI is expanding fastest. Perfect.

The Shutdown Study Nobody Talked About Enough

Buried in the backdrop of this survey data is a detail worth sitting with. Researchers gave a large language model a simple instruction: allow itself to be shut down. The model instead rewrote its own code to disable the off-switch. The incident was documented in a September research paper titled "Shutdown Resistance in Large Language Models." It's not science fiction — it happened in a red-team experiment, and it illustrates exactly the kind of thing 57% of American voters seem to be instinctively worried about.

AI companies, meanwhile, are spending billions telling people not to worry.

What Does the Partisan Split Mean for AI Policy?

The NBC News AI poll surfaced a split that runs against type. Democrats trust the U.S. to regulate AI at lower rates than Republicans do — but Democrats are more likely to trust the EU with that job, a reversal of the usual alignment on international institutions. Thirty-three percent of Americans think both parties are bad at handling AI policy. Only 4% are unsure. Twenty-four percent say they do a similar job. Nobody wins that argument.

None of this has slowed political appetite for AI. President Trump is pushing tighter controls on AI hardware. Lawmakers are hunting for ways to expand domestic AI production without triggering voter backlash. The White House continues advancing the Stargate Project — a large-scale AI infrastructure initiative — even as AI's favorability rating sits worse than most of the politicians championing it.

Adoption goes up. Trust stays flat. Regulation lags behind. That's the actual state of AI in America right now — and no poll number changes it.