What 81,000 People Want from AI: Anthropic's Largest Qualitative Survey

Anthropic has conducted the largest-ever qualitative AI survey, interviewing 80,508 people across 159 countries in 70 languages. Conducted over one week in December 2025 through the 'Anthropic Interviewer' — a Claude-based conversational interview tool deployed to Claude.ai users — it stands as the most multilingual AI opinion study to date.

"Hope and concern coexist as tension within individuals, not between opposing camps." — Anthropic Research Team

The Largest-Ever Candid Conversation About AI

Led by Saffron Huang, the survey's distinction goes beyond scale. While previous AI surveys relied on quantitative multiple-choice formats, this one took a qualitative approach — with AI leading conversations that captured respondents' context and emotions.

Even internally, Anthropic was surprised: "We didn't expect such honest and thoughtful responses." Stories ranged from a patient correctly diagnosed by Claude after nine years of misdiagnosis, to individuals reporting job losses caused by AI.

Nine Visions People Have for AI

Analysis of responses revealed nine distinct visions for AI. Professional excellence topped the list at 18.8%, followed by personal transformation (13.7%) and daily management (13.5%). Time freedom (11.1%), economic independence (9.7%), and societal transformation (9.4%) rounded out the top six.

| Rank | Vision | Share |
|------|--------|-------|
| 1 | Professional Excellence | 18.8% |
| 2 | Personal Transformation | 13.7% |
| 3 | Daily Management | 13.5% |
| 4 | Time Freedom | 11.1% |
| 5 | Economic Independence | 9.7% |
| 6 | Societal Transformation | 9.4% |
| 7 | Entrepreneurship | 8.7% |
| 8 | Learning & Growth | 8.4% |
| 9 | Creative Expression | 5.6% |

Entrepreneurship (8.7%), learning and growth (8.4%), and creative expression (5.6%) filled out the rest. Notably, 81% of respondents said AI was already helping realize their vision — with productivity gains (32.0%) being the most cited benefit, followed by unmet expectations (18.9%) and cognitive partnership (17.2%).

Top Concerns: Hallucination and Jobs

| Rank | Concern | Share |
|------|---------|-------|
| 1 | Reliability/Hallucination | 26.7% |
| 2 | Employment & Economy | 22.3% |
| 3 | Autonomy & Agency | 21.9% |
| 4 | Cognitive Atrophy | 16.3% |
| 5 | Governance | 14.7% |
| 6 | Misinformation | 13.6% |
| 7 | Surveillance/Privacy | 13.1% |
| 8 | Malicious Use | 13.0% |
| 9 | Meaning & Creativity | 11.7% |

In the multi-select concerns section, reliability and hallucination led overwhelmingly at 26.7%. Employment and economy (22.3%) and autonomy and agency (21.9%) followed. The full picture includes cognitive atrophy (16.3%), governance (14.7%), misinformation (13.6%), surveillance and privacy (13.1%), malicious use (13.0%), and meaning and creativity (11.7%).

Surprisingly, existential AI risk came in at just 6.7%. Despite its outsized presence in media narratives, actual users are far more concerned about concrete, everyday problems.

Optimism and Pessimism Live in the Same Person

The survey's most critical finding is that hope and concern don't belong to opposing camps — they coexist within the same individual. Someone who reports dramatically improved work efficiency may simultaneously express anxiety about job security.

This matters because it challenges the framing of AI policy debates. The binary of 'pro-AI vs. anti-AI' fails to reflect reality — and now that claim is backed by data from 80,000 people.

Brazil Most Optimistic, Germany Most Cautious

Overall, 67% of respondents expressed positive sentiment toward AI. The regional breakdown reveals a striking pattern: lower and middle-income countries were more optimistic than wealthier nations. Brazil (71%) and India (70%) showed the highest positivity, while the US (66%) and Germany (64%) were comparatively cautious.

In developed nations where AI has already permeated daily life, firsthand experience with side effects appears to temper enthusiasm. In developing nations, anticipation of AI's opportunities outweighs concern. What 81,000 voices ultimately tell us is this: consensus in the AI era begins not with simple for-or-against, but with navigating complex coexistence.