Most American voters have favorable view of AI — despite media focus on controversies: poll

Not so fast! This poll was produced by an AI advocacy group. I asked Claude (after the story) to evaluate how credible the findings are.

Bottom line: I'm a bit skeptical. Public opinion is more likely slightly against AI. Check out what I found below.

Most American voters have favorable view of AI — despite media focus on controversies: poll

By Thomas Barrabi, NY Post

Published July 20, 2026

More than half of Americans have a favorable view toward artificial intelligence – despite ongoing media firestorms around data centers and potential economic upheaval, according to the results of a recent survey exclusively obtained by The Post.

In a national poll conducted by HarrisX, 56% of voters said they had a favorable view about AI, while 36% had an unfavorable view. Seventy-two percent of respondents agreed that AI could help people spend less time on repetitive tasks, while 71% said they believed AI was the start of a new industrial revolution.

“The American public is actually quite optimistic about the benefits AI will bring to their lives—from medical breakthroughs to economic growth,” said Nathan Leamer, executive director of Build American AI. “Our data shows there is a real opportunity for lawmakers, industry leaders, and innovation advocates to engage Americans on the subjects they care about.”

The poll was commissioned by Build American AI, a nonprofit advocacy group, which argues the results show the perception of a major public backlash against AI development is overblown and that voters are aware of its potential benefits.

Build American AI is an offshoot of pro-tech super PAC Leading The Future, which is bankrolled by industry titans including OpenAI President Greg Brockman, Palantir co-founder Joe Lonsdale and Andreessen Horowitz.

For example, 61% of respondents agreed that AI could “help cure diseases, discover new medicines, and accelerate science,” while 64% agreed that AI would “make the US more competitive against China.”

Elsewhere, 60% of voters said they felt AI could help America reindustrialize and build more products domestically.

The divide was narrower when it comes to potential safety risks associated with AI development, with 53% of voters saying America should move quickly to build AI despite the risks because foreign nations are pursuing it, compared to 47% who disagreed.

AI data centers in particular have been a source of major consternation, with critics alleging they are harmful to local environments and consume huge amounts of energy to the public’s detriment. In New York, Gov. Kathy Hochul recently signed an executive order imposing a one-year moratorium on data center construction within the state.

That move generated sharp pushback from both local business leaders and President Trump, the latter of whom called it a “terrible decision.”

Some top AI executives, including Anthropic CEO Dario Amodei, have further stoked tensions with warnings that the industry could wipe out white-collar and early-career jobs at AI becomes more advanced and takes over more work from humans.

At the same time, the White House and other tech leaders insist that advanced AI, such as Anthropic’s Fable model, are critical to national security.

HarrisX conducted interviews with 2011 registered voters across the country in July 2026.

Short answer: yes, some skepticism is warranted — but that doesn't mean the poll is wrong, just that it needs to be read carefully.

A few reasons for caution:

1. It's an outlier relative to most other 2026 polling. Nearly every other major survey this year — NBC News/Hart-POS (Feb-March), Fox News, the July Harvard CAPS/Harris poll, Pew — has found Americans either evenly split or net negative on AI, often by wide margins (NBC had it at -20 net favorable, worse than ICE or either political party; Harvard/Harris found two-thirds want to slow AI development down). Data for Progress found opinion "closely divided" (48% favorable/46% unfavorable). A poll landing at "most voters favorable" cuts against that trend, which is itself a reason to look closely at how it was fielded.

2. Industry-funded pollsters have a track record here. Groups like the American Edge Project (an AI/tech industry coalition) and pro-AI PACs like Leading the Future have released a string of polls this year consistently finding voters supportive of AI and unbothered by regulation — findings that conveniently track their policy goals (less regulation, more data center buildout). That doesn't make the numbers fake, but sponsor-funded polls are more prone to favorable framing: question wording, order effects, which subgroups get emphasized, and what's highlighted in the press release.

3. What to actually check, rather than dismissing it outright:

4. Who fielded it (a named pollster like Morning Consult or Ipsos is more reassuring than an in-house or unnamed firm)

5. The exact question wording — "favorable/unfavorable" vs. "AI will change your life a lot" vs. "AI is good for society" can produce very different numbers on the same underlying attitudes

6. Sample and methodology (registered voters vs. likely voters vs. general population, sample size, field dates)

7. Whether the toplines match the crosstabs, or whether a favorable topline is being driven by a subgroup (heavy AI users, younger men) while other segments remain skeptical — that pattern shows up even in neutral polls like Data for Progress's

So: don't assume it's bunk, but treat "commissioned by a pro-AI group" as a flag to check methodology and wording before taking the headline number at face value, especially since it diverges from most other polling this year.

What's more credible than a generic "industry poll":

• HarrisX is a real, established pollster (does work for Axios, Fortune, and others across the spectrum) — this isn't an unnamed in-house shop. A large sample (2,011 registered voters) is also solid.

• The specific numbers are traceable to an identified pollster and sponsor, not just a press release with no attribution.

What still warrants skepticism:

1. It's a significant outlier for the same month. The July 2026 Harvard CAPS/Harris poll — released just days earlier — found two-thirds of voters want to slow down AI development. NBC's spring poll had AI at net -20 favorability. Data for Progress in February found opinion essentially tied (48-46). This poll's 56-36 favorable is well outside that range. That gap needs an explanation beyond "the public is actually optimistic" — it could reflect real movement, but it's more likely a signal to look at wording.

2. The question wording pattern is a tell. Notice the framing on the headline stats: "AI could help cure diseases," "AI is the start of a new industrial revolution," "AI would make the US more competitive against China." These are all agree/disagree with a positive vague statement questions — a format known to produce acquiescence bias (people tend to agree with plausible-sounding statements, especially patriotic or aspirational ones). That's different from asking a neutral "do you have a favorable or unfavorable view of AI" without a preceding frame, or asking about specific harms.

3. The one question closest to a real tradeoff is telling. The only item in the piece that pits AI's benefits against its risks directly — "move quickly despite risks because other countries are doing it" — splits 53-47, far closer than the flashy 60-70% agreement numbers. When risk is put on the table explicitly, the gap nearly disappears. That's a strong hint the other questions were framed to elicit favorable responses rather than measure genuine sentiment.

4. Selective release. The Post calls this "exclusively obtained," meaning Build American AI chose what to hand over — likely the best-performing findings from a longer questionnaire, not full toplines or crosstabs. Without the actual instrument (exact wording, question order, full result set), you're seeing the sponsor's highlight reel.

5. Timing is strategic, not incidental. This dropped right after Hochul's data-center moratorium and Amodei's job-loss warnings — i.e., during a negative news cycle for AI. That's consistent with an advocacy communications push (which is exactly what Build American AI is) rather than an independent research release.

Bottom line: The pollster's credentials make outright fabrication unlikely, but the question framing looks built to produce favorable-sounding toplines, the results are a clear outlier against concurrent independent polling (including one from the same week), and you're only seeing what the sponsor chose to share. Worth treating the 56-36 headline as "Build American AI's most favorable framing of a larger survey" rather than a clean read on where the public actually stands — and weighing it against NBC, Harvard/Harris, and Data for Progress rather than in isolation.