Poll: Most Americans say AI is hurting, not helping, K-12 and college students
A new national survey finds a majority of U.S. adults now believe schools that ban AI serve students better than schools that embrace it, a sharp swing from earlier this year.
What happened: An NBC News Decision Desk Poll powered by SurveyMonkey, surveying 7,105 adults in late August, found 53% believe AI does more harm than good in K-12 schools, versus 27% who see more good than harm. Higher education numbers were nearly identical: 54% harm, 24% good. Most notably, 56% now say a school that bans AI will better prepare students for the future, up 9 points since June, while only 44% back schools that integrate it, down 9 points.
Why it matters: This is a real reversal in public mood within a few months, not a static split. It's happening as districts make binding decisions right now, most visibly New York City, which just announced it will restrict student AI use until high school with only limited allowances after that. Public opinion is shifting toward caution just as the technology becomes cheaper and more embedded in everyday apps students already use.
How it works, plainly: Respondents weren't uniformly anti-AI; they drew lines by task. Using AI to proofread or edit was seen as acceptable by 49%, the most tolerated use. Getting help researching or brainstorming was OK to 41%, and using it to outline or organize ideas was fine to 38%. But only 7% said it was acceptable for AI to actually write or generate content for a student, and 31% said no classroom AI use at all was acceptable.
The rollout: Policy is diverging in real time. NYC's new limits reflect the more restrictive public mood. Meanwhile first lady Melania Trump told NBC News she wants students getting hands-on AI experience now, warning that the U.S. risks falling behind if it doesn't teach the technology early, while also urging vigilance. That tension, between guarding against harm and not leaving kids unprepared, is exactly what the poll numbers capture, and it's unresolved.
