Jonathon's AI Wiki

Pew Research — How Americans View AI and Its Impact (Sept 2025)

9 min read

Source: pew-americans-view-ai-2025-09-2026-05-19.md — Pew Research Center (Brian Kennedy, Eileen Yam, Emma Kikuchi, Isabelle Pula, Javier Fuentes). Published September 17, 2025. Fielded June 9-15, 2025. n=5,023 U.S. adults via American Trends Panel (ATP), national random address-based sample, online and telephone with live interviewer. Weighted to be representative of the U.S. adult population by gender, race, ethnicity, partisan affiliation, education, and presidential vote.

A nationally representative survey of 5,023 American adults finds that nearly all U.S. adults (95%) have heard at least a little about AI — but the dominant emotional posture is concern, not excitement. Half say they are more concerned than excited, more than half rate AI’s societal risks as high, and majorities expect AI to worsen people’s ability to think creatively and form meaningful relationships. Awareness has reached near-saturation; favorable sentiment has not.

Key Takeaways

  • Near-universal awareness, net-negative affect. 95% of U.S. adults have heard at least a little about AI. Yet 50% say they are more concerned than excited about its increased use in daily life — up from 37% in 2021. Only 10% are more excited than concerned; 38% say they are equally excited and concerned.
  • Risk perception outpaces benefit perception. 57% rate the societal risks of AI as high; only 25% say the benefits are high. When asked to explain their high-risk rating in their own words, the most common concern cited was AI weakening human skills and connections.
  • Creative thinking — pessimism dominates. 53% say AI will worsen people’s ability to think creatively, versus 16% who say it will improve this. Another 16% expect no change; 16-20% across skill questions say they are unsure.
  • Meaningful relationships — even starker. 50% say AI will worsen people’s ability to form meaningful relationships, versus only 5% who expect improvement. One-quarter say it will make this neither better nor worse.
  • Problem-solving — relatively more optimistic, still net-negative. 29% say AI will make people better at problem-solving; 38% say worse. This is the least pessimistic of the three skills tested.
  • Control deficit is widespread. About six-in-ten want more control over how AI is used in their lives; only 17% are comfortable with their current level of control. Yet nearly three-quarters say they would be willing to let AI assist at least a little with day-to-day tasks — willingness and anxiety coexist.
  • Detection confidence gap. 76% say it is extremely or very important to be able to tell if pictures, videos, and text were made by AI or by people. But 53% say they are not too or not at all confident they can detect AI-generated content — a majority who hold high stakes for detection but doubt their own ability to do it.
  • Personal vs. analytical domains. Majorities support AI for analytical roles: forecasting weather (74%), searching for financial crimes (70%), searching for government benefits fraud (70%), developing new medicines (66%), and identifying crime suspects (61%). AI gets no role, per two-thirds of respondents, in matchmaking; and 73% say it should play no role in advising people about faith in God.
  • Young adults diverge on awareness but converge on pessimism. 62% of adults under 30 have heard or read a lot about AI, versus 32% of those 65+. The awareness gap has grown since 2022. Yet majorities of adults under 30 still say AI will worsen creative thinking (61%) and meaningful relationships (58%) — slightly higher than older cohorts (~40%), not lower. Heavy awareness does not translate to favorable outlook.

Survey Design

  • Sample: n=5,023 U.S. adults, recruited from the American Trends Panel — a nationally representative, probability-based online panel built from random address-based sampling of U.S. residential addresses. Every U.S. adult has a known, non-zero probability of selection.
  • Fielding window: June 9-15, 2025 (six days).
  • Mode: Primarily online self-administered; telephone with live interviewer available for panel members without internet access.
  • Weighting: Representative of the U.S. adult population by gender, race/ethnicity, partisan affiliation, education, and presidential vote (among voters).
  • What this study measures: Attitudes, perceptions, and stated willingness — not actual behavior. The survey captures how Americans say they feel and what they say they would do; it does not track whether they use AI tools, how often, or with what outcome.
  • Limitations:
    • Fielded six months before the September 2025 publication date (June 2025 data); public sentiment may have shifted by the time the report circulated.
    • Self-reported “concern” and “excitement” are highly sensitive to question wording and order effects; Pew’s replication of these items from 2021 enables trend comparison but does not eliminate framing artifacts.
    • The survey does not segment by AI tool type (generative AI vs. AI in recommendations vs. AI in hiring), so “AI” is conflated across very different product surfaces.
    • Related Pew releases in fall 2025 (October: “How People Around the World View AI”; November: “Republicans, Democrats now equally concerned”) suggest the sentiment picture was evolving rapidly through this period — point-in-time survey may understate movement.

Why This Matters for AI Strategy and Marketing

This survey is the public-perception layer beneath the AI adoption and AI-search data the rest of this wiki tracks. The headline read for practitioners building AI-driven brand experiences or AI search visibility: the audience is large and aware but uncertain and anxious, not enthusiastic.

Three implications translate directly to practice:

1. Do not lead with “powered by AI.” The dominant emotion in the audience is concern (50% more concerned than excited). Framing an experience as “AI-driven” activates that concern before the user has experienced any benefit. The smarter frame is outcome-first — lead with what the product does, not how it does it. Reserve the “AI” label for contexts where transparency is a trust signal (healthcare, financial fraud detection, medicine), not a generic badge.^[inferred]

2. Transparency and human-in-the-loop signalling are load-bearing. 76% of respondents say it is extremely or very important to be able to distinguish AI-generated from human-made content — yet 53% don’t trust their own ability to do it. That combination produces anxiety, not just skepticism. Products that make the human-in-the-loop visible (named reviewer, editorial date, attribution) reduce that anxiety more reliably than feature marketing. This is the design principle beneath the Hallmark design skill’s anti-AI-aesthetic posture and the same logic Hormozi’s AI system uses when it separates voice (human data advantage) from AI execution.^[inferred]

3. Connect to the AI-SEO citation context. The AI SEO hub and its underlying research cluster show what happens when AI search systems surface content: citation patterns are driven by authority, structure, and topical depth. But the user sitting at the other end of those AI answers is — per this Pew data — uncertain, skeptical of AI-generated content, and craving the ability to verify. Content that visibly signals human authorship, expert attribution, and citable primary sources performs both layers simultaneously: it earns AI citations AND is trustworthy to the human who receives them.^[inferred]

The Gen Z resistance data in Gen Z AI Resistance and the Pew survey are complementary cuts at the same sentiment reality. Gen Z (ages under 30) in the Pew survey show higher awareness than older cohorts but similar or higher pessimism — consistent with the Gen Z article’s “familiarity increases concern, not excitement” finding. For marketing teams deploying AI, the combined read is: the more your audience knows about AI, the less likely naive AI-boosterism will land.

Open Questions

  • How has sentiment moved since June 2025? Three related Pew releases in fall 2025 (October global survey, November partisan-alignment piece, October AI-in-news piece) suggest rapid evolution through the period. The June 2025 Pew baseline may already be outdated for strategy decisions made in mid-2026.^[ambiguous]
  • Does concern cluster around specific AI product surfaces, or is it generalized? The survey uses “AI” as an undifferentiated term. Concern about AI in hiring may be far higher than concern about AI in weather forecasting (where 74% support a role for AI) — but the survey cannot separate these. A follow-up that segments by product type would have much higher strategy value.
  • Does the 50% “more concerned than excited” figure shift when respondents are asked about AI tools they personally use versus AI in society broadly? Framing effects around personal use versus societal impact consistently produce divergent readings in technology-attitude surveys.^[ambiguous]
  • How does this Pew data reconcile with the WalkMe adoption survey’s finding that executives are far more enthusiastic than employees? The WalkMe State of Digital Adoption 2026 shows executive-employee adoption gaps; it would be worth knowing whether executive AI sentiment also diverges from the general-public baseline captured here.

Try It

  1. Audit your AI messaging. Grep your website, email sequences, and landing pages for the phrase “powered by AI” or “AI-generated.” For each instance, ask: does surfacing the AI label here reduce anxiety or increase it for a user who is more concerned than excited? Replace outcome-neutral AI badges with outcome-first copy.
  2. Add visible attribution to AI-assisted content. If your content pipeline uses AI (writing, images, summaries), add a brief attribution line naming the human editor and review date. 76% of Americans say verifiability matters enormously — a simple byline satisfies that need without burying the efficiency gains of AI-assisted production.
  3. Segment your audience by AI sentiment before personalizing AI experiences. Use an opt-in preference center or behavioral signal (did the user engage with an AI chatbot vs. navigate to a human contact form?) to route AI-skeptical users to human-forward experiences. The survey makes clear that the “more concerned” cohort is a majority — design for them as the default, not the exception.
  4. Use this Pew data to brief stakeholders on realistic AI adoption curves. If an internal AI initiative is being sold with “everyone is excited about AI,” the 50%/10%/38% split is corrective evidence from a large, nationally representative sample. Anchor adoption roadmaps to where people actually are emotionally, not where vendors say they are.
  5. Cross-reference with AI-search content strategy. Per the AI SEO hub, AI search engines are pulling content from authoritative, well-structured sources. Per this Pew survey, the humans receiving those AI-mediated answers want verifiability and human attribution. Write for both: structured for citation, signed for trust.

Graph View

Backlinks