Public use of generative AI grows, but trust and comfort with news applications remain low

A global survey conducted in 2025 by the Reuters Institute and the University of Oxford found a significant increase in public awareness and use of generative artificial intelligence, with weekly usage of tools such as ChatGPT rising from 18% in 2024 to 34% in 2025.

The study, conducted in Argentina, Denmark, France, Japan, the United Kingdom and the United States, also examined how generative AI is used in news, search and broader societal contexts.

ChatGPT remained the most used and recognized tool, with 22% of respondents reporting weekly use. Awareness of generative AI tools increased across all countries, with 90% of participants aware of at least one tool, up from 78% the previous year. Younger respondents, particularly those aged 18 to 24, showed the highest levels of adoption.

Information-seeking became the leading use case for generative AI, surpassing content creation. Weekly usage for information doubled to 24%, while use for media creation increased to 21%. News-related applications of AI remained limited, with only 6% of respondents using AI systems to get news each week. The highest usage for news occurred in Argentina and the United States.

Respondents expressed mixed views on the reliability of AI-generated search results. While 54% reported seeing AI-generated answers in searches, only half of those said they trusted the information. Click-through rates on source links were low, and many users reported verifying AI answers with traditional sources, particularly in areas like health and politics.

The survey also found that many people believe AI is already widely used in various sectors.

More than half said they believe AI is often used in news media, with even higher figures for search and social media companies. However, optimism about AI’s impact varied by sector. Respondents were more positive about AI in healthcare, science and search engines but were more pessimistic regarding its use in news media, government and politics.

In journalism specifically, a significant gap exists between comfort levels with human-produced and AI-produced content. Only 12% of respondents said they were comfortable with fully AI-generated news, compared to 62% who preferred human-made content. Acceptance increased to 43% when a human led the process with AI assistance.

Advertisement

People were more accepting of back-end uses of AI, such as grammar correction and translation, than of front-facing applications like AI-generated images or artificial presenters. Only one-third of respondents believed journalists regularly check AI-generated content before publication, and 43% expected large differences between news outlets in how responsibly they use AI.

Although a majority of people consume news daily, only 19% reported seeing AI labeling on news stories each day. AI summaries and chatbots were the most frequently recognized features, but 60% of respondents said they did not regularly encounter AI features in news platforms.

The report also highlighted national differences in perception. Respondents in Japan and Argentina were generally more optimistic about AI’s societal impact, while participants in the United Kingdom and the United States expressed more skepticism.

The findings are part of the Reuters Institute’s ongoing research into public attitudes toward AI in journalism and society, supported by the National Center for AI in Society and the Thomson Reuters Foundation.