Who trusts AI?

AI tools like ChatGPT and Google Gemini have exploded in popularity over the last couple years. But while their adoption has undoubtably grown at a staggering speed, less is known about public attitudes towards the trustworthiness of these systems.

Since the launch of ChatGPT in late 2022, generative artificial intelligence (AI) tools have rapid adoption. Estimates of use vary. A recent Federal Reserve backed survey found 55% of US adults use AI tools, while other sources report lower use: figures from Microsoft’s AI for Good Lab suggest use under 30% in Denmark, around 35% in the UK, and over 60% in Singapore.

In any case, adoption of AI tools – digitally distributed and low cost – has been far faster than for past general-purpose technologies. AI usage runs years ahead of the early adoption paths of the internet or personal computer, for example.

Yet while people have adopted AI tools quickly, how they feel about them is more ambivalent. Across the world, survey evidence suggests a mixture of anxiety and excitement, with attitudes split sharply by country, age and application.

Opinion polling by Pew Research Centre, a US-based think tank, shows that concern outweighs excitement almost everywhere. Respondents in every country surveyed, except Israel and South Korea, reported feeling more concerned than excited about increased use of AI in daily life. Common worries include impacts on human skills, jobs and privacy, as well as the spread of inaccurate information.

Chart 1. Concern and excitement about AI

Source: Pew Research, Spring Global Attitudes Survey 2025
Notes: Proportion who say the increased use of artificial intelligence in daily life makes them feel more concerned or excited.

Attitudes toward AI vary widely across countries. The United States emerges as one of the most sceptical: 50% of Americans report feeling more concerned than excited, compared with just 10% who feel more excited than concerned. Other English-speaking countries follow closely, a trend found in other analysis also.

At the other end of the spectrum lies South Korea, India, Israel and Nigeria. There appears to be less uniting the AI ‘boosters’ than the sceptics: these countries vary in their income levels, geography and political context.

Why are some countries more optimistic than others?

One striking relationship is with income. Richer countries are consistently more pessimistic about AI’s impact.

Chart 2. AI attitude and GDP per capita

Source: Artificial Intelligence Index Report 2025 (Stanford/IPSOS), World Bank, author’s calculations
Notes: Proportion who say the increased use of artificial intelligence in daily life makes them feel more concerned or excited.

In polling conducted by Ipsos for Stanford’s Artificial Intelligence Index 2025, every country with a GDP per capita below $36,000 – around the level of Malaysia or Chile, after adjusting for differences in local prices – expects AI to deliver more benefits than drawbacks. In contrast, richer countries, including the United States and European economies such as Switzerland, the Netherlands and Norway, are the most cautious. Singapore stands out as an exception: both wealthy and strongly optimistic about AI.

This income-AI link is consistent across independent analyses. A recent cross-country study of 14,000 workers in 31 countries finds that higher levels of socio-economic development are associated with less optimistic expectations about AI’s economic impact. Though hard to test, the paper posits a few possible explanations:

Leapfrogging instead of displacement. In lower-income countries, AI can be seen as a way to modernise industries and compensate for skill gaps. In richer countries, fears of AI displacing workers are more salient.
Exposure and awareness of risk. Workers in more developed countries tend to have higher experience with, and knowledge of, information technologies which can heighten awareness of dangers.
Skills composition. While previous waves of automation have disrupted lower-skill employment such as manufacturing, the generative AI wave has automated service tasks that are more common in more developed countries.

How do attitudes vary with gender and age?

Within countries, attitudes toward AI also vary significantly with age. In every country surveyed by Pew, adults aged 50 and over are more likely to report concern about AI than those aged 18–34 (Chart 3).

Chart 3. More concerned than excited, by age

Source: Pew Research, Spring Global Attitudes Survey 2025
Notes: Proportion who say the increased use of artificial intelligence in daily life makes them feel more concerned or excited.

In the UK, for example, 46% of those aged 50+ say they feel more concerned than excited about AI, compared with 31% among younger adults. In Greece, the gap is widest: just 18% of those aged 18–34 report feeling more concerned than excited, compared with 59% of those aged 50 and over.

Gender differences are also common. Women consistently report higher levels of AI concern than men. Pew polling findsthat American women are less likely to report a positive effect of AI technology on society: 42% versus 54% and less comfortable with every AI application polled, from medical diagnosis to household tasks.

Attitudes towards AI also depend on how and where it is applied. People are much more prepared to trust AI for lower-stakes applications in leisure, entertainment and hospitality than for financial or health decisions. Just 15% of Americans surveyed by YouGov say they trust AI applications for financial services, compared with 48% who responded ‘not much’ or ‘not at all’. The gap between trust and distrust – net trust, shown below – is negative for every category surveyed except leisure and entertainment, which is perfectly balanced. It may be no surprise then that AI media, particularly music and shortform video, has had a wide reach, topping charts and fooling listeners.

Chart 4. Trust in AI by sector (United States)

Source: YouGov, December 2025, author’s calculations
Notes: Net trust (trust a lot/somewhat less do not trust much/at all) by sector.

Where next?

The public caution and distrust are a sharp contrast to the enthusiasm visible in financial markets. AI startups, led by OpenAI and Anthropic are expected to go public with initial public offerings (IPOs) valued in the hundreds of billions, while established tech giants and ‘megascalers’ are expected to spend upwards of $300 billion on AI development and datacentres. AI products’ rapid development and spread, supported by a mass-mobilisation of capital, are increasingly in tension with distrust and concern among the public.

AI has also emerged as a fault-line in western politics:

United States. Democrat senators, including Bernie Sanders, Edward J. Markey, and Elizabeth Warren have pushed for a moratorium on AI data centre construction. At the same time, a bi-partisan coalition, from Steve Bannon to Warren, has mobilised against efforts to block US states from regulating AI.
United Kingdom. Both the current Labour government and the preceding Conservative administrations have courted AI development, with AI accelerators and reforms. In contrast, the Green Party has struck a more cautious tone, prioritising a Digital Bill of Rights, in their last manifesto.
European Union. At the EU level, public caution AI has translated most clearly into regulation with a system-wide approach through the European Union’s AI Act.

It remains to be seen which will adjust first: public attitudes toward AI, the scale of market enthusiasm, or the political context around its deployment. Which force gives way will determine the future and integration of the 2020s most disruptive technology.

Where can I find out more?

Attitudes to AI. A national survey of attitudes to AI in the UK, backed by the Ada Lovelace Institute and the Alan Turing Institute.
Both sides of the aisle hate the AI moratorium. Reporting from The Verge on US AI regulation efforts.
Artificial Intelligence, Pew polling. Survey evidence on attitudes to AI around the world.
Artificial Intelligence, YouGov UK. UK polling evidence.