AI Trust: 87% in China, 32% in the US. Why the Gap?

87% of surveyed Chinese citizens say they trust artificial intelligence. In the United States, that figure drops to 32%. This is not a statistical nuance: it is a 55 percentage point difference that raises uncomfortable questions about which cultural, political, and economic factors shape a society's relationship with technology. The data comes from the Edelman Flash Poll Trust and Artificial Intelligence at a Crossroads, published in November 2025 and circulating again this week on Hacker News, generating considerable debate.

The report, based on surveys of citizens across several countries, does more than capture the headline-grabbing China versus US contrast. It points to a broader pattern: countries with higher general institutional trust also tend to show higher trust in AI, suggesting the issue is not the technology itself, but the context in which it is deployed.

The institutional factor: trusting those who deploy AI

One of the most solid readings of the report is that trust in AI is not measured in the abstract. Citizens trust, or distrust, the actors who develop and regulate that AI: companies, governments, technical bodies. In China, where official narrative frames AI as a tool for national prosperity and the State takes an active role in its rollout, the frame of reference differs fundamentally from the American one.

In the US, by contrast, AI reaches public opinion after years of tech scandals, privacy breaches, algorithmic bias, and disinformation campaigns led by private companies. Distrust in Big Tech transfers, almost automatically, to any technology those same companies lead. The 32% figure does not say Americans are more technically critical of AI; it says they trust less in those who build it.

The role of exposure and familiarity

Another thread emerging from the report is the correlation between daily exposure to AI tools and confidence levels. Countries where AI integrated earlier and more deeply into everyday services—payments, transport, health, customer support—tend to show higher figures. Familiarity reduces diffuse fear, though not necessarily critical scrutiny.

This has direct implications for the AI development ecosystem: mass adoption of assistants, conversational interfaces, and autonomous agents is not just a commercial goal, but also a variable that affects public perception of the technology in the medium term. It is no coincidence that several European countries, with more fragmented adoption and more visible regulation, appear in middle positions in the rankings.

What does "trust" actually measure here?

It is worth being precise about what the term encompasses in these types of surveys. Edelman measures stated perceptions, not actual behaviour. A person can answer that they "trust" AI in the sense that they believe it will be useful or will not harm them, without that implying they understand how it works, have read a privacy policy, or demand transparency from those deploying it.

That ambiguity matters. An 87% trust level in China may reflect genuine optimism, but also the absence of a public space to express doubts. 32% in the US may reflect both informed skepticism and reactive rejection fueled by negative news cycles. The numbers do not distinguish between these things, and headlines typically do not either.

For whom is this data relevant

For those deploying solutions based on Claude or other models in enterprise or end-user contexts, these numbers are not academic. Social acceptance of an AI product varies enormously depending on market, sector, and user profile. Building interfaces that clearly communicate what the system does, what data it uses, and who is accountable for its decisions remains one of the most underestimated factors in product design.

The Edelman report is from November 2025, but its conclusions remain pertinent: the trust gap between markets does not close with better models, but with better narratives, better regulatory frameworks, and above all, better transparency practices.

Editor's view: The difference between 87% and 32% says more about each country's institutional ecosystems than about AI itself. Any adoption strategy that ignores that context is probably measuring the wrong problem.