Public trust in AI-generated news remains low despite rising usage

A comprehensive survey released on October 7, 2025, by the Reuters Institute for the Study of Journalism at Oxford University documents significant shifts in how people interact with generative artificial intelligence systems while maintaining persistent skepticism about their role in journalism.

The research, conducted between June 5 and July 15, 2025, across Argentina, Denmark, France, Japan, the United Kingdom, and the United States, found that public awareness of artificial intelligence tools reached 90%, with weekly usage nearly doubling from 18% to 34% in just one year. ChatGPT dominated recognition at 73% awareness, followed by Google Gemini at 50% and Meta AI at 46%.

Search experiences transform information retrieval

The data showed that 54% of respondents encountered AI-generated search answers in the past week, with the highest exposure in Argentina at 70%, followed by the UK at 64% and the USA at 61%. France recorded the lowest figure at 29%, likely because Google had not deployed AI overview features in that market during the survey period.

Among those who saw AI-generated search responses, one-third reported clicking through to source links always or often, while 28% rarely or never clicked through. Trust in these AI answers reached 50% among those who encountered them, with younger adults showing slightly higher confidence than older users.

Click-through behavior appeared correlated with trust levels. Respondents who trusted AI search responses proved more likely to click through to underlying sources at 46%, compared to just 20% among those who distrusted the answers. This pattern suggests links serve to gather additional information rather than validate AI responses.

Information-seeking overtakes media creation

Weekly use of generative AI for information retrieval more than doubled from 11% to 24%, surpassing media creation activities, which grew 7 percentage points to 21%. The shift marked information-seeking as the primary application for these technologies.

Specific information-retrieval tasks showed strong growth, with users employing AI to answer factual questions increasing from 6% to 11%, and those using it for advice rising to the same level. A new survey category, "researching a topic," attracted 15% of respondents.

Image generation drove growth in media creation, rising from 5% to 9% weekly usage, while video generation remained at 3% and audio at 2%. Programming and coding applications stayed flat despite improved systems, suggesting early adopters had already embraced these capabilities in 2024.

The proportion using generative AI for news doubled from 3% to 6%, though this increase concentrated in Japan and Argentina, with other markets showing little change. Among AI news users, 54% sought the latest news, while 47% asked follow-up questions about stories.

Age differences emerged in news consumption patterns. Younger users aged 18-24 employed AI to make news stories easier to understand at 48%, compared to 27% among those 55 and older. Older groups focused more on obtaining the latest news at 59% versus 39% for younger respondents.

Brand recognition concentrates around major platforms

ChatGPT maintained dominance with 22% weekly usage across the six countries, followed by Google Gemini at 11%, Meta AI at 9%, and Microsoft Copilot at 6%. The user base for every major system roughly doubled since May 2024.

Despite rapid growth, most people remained occasional rather than regular users. Many reported monthly usage or having tried systems once or twice, with substantial shares never having used or heard of various platforms.

Generative AI use skewed younger, with 59% of 18-24 year-olds reporting weekly usage compared to 20% among those 55 and over. However, age gaps narrowed for Microsoft Copilot, Meta AI, and Grok, likely due to their integration into widely-used products.

Trust levels concentrated among major brands. ChatGPT led at 29%, followed by Google Gemini at 18%, Microsoft Copilot at 12%, and Meta AI at 12%. Most other brands registered trust below 10%, primarily due to low awareness.

Country-level trust patterns varied significantly. ChatGPT enjoyed positive trust in most markets except the UK, where more people distrusted than trusted it. Argentina showed no systems with negative trust scores, while DeepSeek registered negative trust in most countries except Argentina.

Perceptions of artificial intelligence across sectors

Public perception indicated generative AI had become widespread, with 41% believing it was used "always or often" across different sectors. This figure far exceeded the 15% who said it was used "rarely or never."

AI integration in search and social media drove particularly high awareness, with 67% believing search engine companies used AI frequently and 68% for social media companies. News media registered 51% for perceived frequent use.

Expectations about AI's impact on user experiences varied by sector. On average, 29% expressed optimism and 22% pessimism about AI improving their interactions with different institutions. Optimism exceeded pessimism for healthcare, science, and search engines, while pessimism dominated for news media, government, and especially politicians and political parties.

When comparing personal versus societal impact, optimists outnumbered pessimists in four of six countries regarding individual lives. However, when considering societal effects, pessimists dominated in three countries, including the United States.

Female respondents showed significantly lower expectations than males, being less likely to anticipate positive personal or societal outcomes from generative AI. Younger people under 35 tended toward greater optimism on both personal and societal levels.

News production faces persistent skepticism

The survey revealed a substantial "comfort gap" between AI-driven and human-led news production. Only 12% felt comfortable with entirely AI-generated news, rising to 21% with human oversight, 43% when humans led with AI assistance, and 62% for entirely human-produced journalism.

This gap appeared across all six countries, though it measured smaller in Japan and Argentina compared to Denmark and the UK. The gap widened slightly from 2024, with comfort in entirely human journalism increasing 4 percentage points.

Comfort levels varied by journalistic task. Back-end applications received greater acceptance, with 55% comfortable with AI editing spelling and grammar and 53% accepting translation services. Front-facing uses drew more resistance, including rewriting content for different audiences at 30%, creating realistic images when photographs were unavailable at 26%, and generating artificial presenters or authors at 19%.

Perceived prevalence of AI use in newsrooms increased across all tasks by at least 3 percentage points from 2024. The proportion believing journalists "always or often" used AI for grammar editing rose 8 percentage points from 43% to 51%.

Public perception aligned with preferences, showing people believed journalists used AI most frequently for tasks they found most acceptable. Grammar editing ranked highest for both comfort at 55% and perceived usage at 51%, while artificial presenters scored lowest on both measures at 19% comfort and 20% perceived usage.

Expectations about news quality remain mixed

Respondents held differentiated views about how AI would affect news qualities. People expected AI would make news cheaper to produce, with a +39 net score, and more up to date at +22. However, they anticipated declines in transparency at -8 and trustworthiness at -19.

These views hardened slightly from 2024, with some net scores increasing but none decreasing. The pattern suggested people believed generative AI would primarily benefit publishers rather than users.

Japan and Argentina showed more positive assessments, with people expecting AI would improve most news qualities. European countries and the USA demonstrated more differentiated views, while UK respondents thought AI would worsen many aspects of news.

Only 33% believed journalists "always or often" checked AI outputs before publication. This figure varied by country, measuring higher in Japan at 42% and Argentina at 44%, but lower in the UK at 25%. Compared to 2024, the USA showed a slight decrease of 3 percentage points, while Denmark, Japan, and Argentina recorded increases.

Trust in news strongly correlated with beliefs about checking procedures. Among those who strongly trusted news, 57% thought journalists regularly checked AI outputs, dropping to just 19% among strong distrusters.

Respondents expected large differences in how responsibly different outlets would use generative AI, with 43% anticipating very or somewhat large differences compared to 28% expecting small differences. France proved the exception, showing roughly equal proportions at 35% and 34% respectively.

Limited visibility of audience-facing features

Most respondents at 60% reported not regularly seeing AI-powered, audience-facing features like bullet-point summaries and chatbots on news sites. Among those who did encounter such features, AI summaries proved most common at 19%, followed by chatbots at 16%.

AI labeling appeared infrequently relative to daily news consumption. Only 19% saw AI labels daily and 28% weekly, despite 77% reporting daily news use. About 15% often or always suspected AI use without labeling, with Argentina highest at 30% and European countries around 10%.

The survey data emerged from approximately 2,000 respondents in each country, assembled using nationally representative quotas for age, gender, region, and political leaning. The research represented a follow-up to a 2024 survey using the same methodology.

Online samples tend to under-represent people who are not online, typically those who are older, less affluent, and have limited formal education. Additionally, online survey panels tend to over-represent well-educated, socially and politically active people.

Why this matters for marketers

The research carries significant implications for the marketing community, particularly as AI transforms search advertising. Microsoft research published in June 2025 documented that conversational search interactions compressed traditional customer journeys by 40% fewer touchpoints, with purchase rates improving 53% following AI-powered search tool interactions.

Google's integration of advertisements into AI Overviews, announced for mobile users in the United States, represents a fundamental shift in how people encounter commercial messages during information-seeking. The company's head of search revealed that AI Mode users issue queries 2-3 times longer than traditional search, enabling better targeted advertising.

Platform automation continues expanding, with Google announcing the removal of language targeting settings from search campaigns by the end of 2025. Language will be automatically detected using artificial intelligence, removing manual configuration requirements.

The publisher compensation landscape faces disruption as AI platforms increasingly extract value from content. IAB Europe released technical standards requiring artificial intelligence platforms to compensate publishers for content ingestion, addressing documented traffic disruptions where referrals from AI platforms increased 357% year-over-year in June 2025.

For marketing professionals navigating this transformation, the Interactive Advertising Bureau released guidance on September 3, 2025, providing structured frameworks for understanding AI applications across campaign lifecycles. The comprehensive map organizes use cases into six categories with maturity indicators distinguishing established from emerging technologies.

Alternative AI platforms like Perplexity disclosed visual implementations of advertising programs in November 2024, illustrating how sponsored content integrates with search results while maintaining answer integrity through sponsored follow-up questions and paid media positioned adjacent to answer content.

The trust gap documented in the research suggests marketers must carefully consider transparency and user experience when deploying AI-powered features. While technical capabilities advance rapidly, public skepticism about AI-generated content, particularly in high-stakes domains like news and health, requires measured approaches that prioritize human oversight and clear disclosure.

Timeline

• April 9-30, 2024: Original survey conducted across six countries documenting early AI adoption patterns
• May 6, 2025: Google announces AI Max for Search campaigns promising 14% conversion improvements
• June 5, 2025: Microsoft releases comprehensive whitepaper on AI search transformation
• June 5-July 15, 2025: Follow-up survey fieldwork conducted by YouGov across Argentina, Denmark, France, Japan, the UK, and the USA
• June 25, 2025: Google's search head announces AI Mode advertising rollout
• September 3, 2025: IAB releases AI use case map for advertising professionals
• September 2025: IAB Europe unveils framework for AI publisher compensation
• October 7, 2025: Reuters Institute publishes comprehensive research findings

Summary

Who: Researchers Dr. Felix Simon, Prof. Rasmus Kleis Nielsen, and Dr. Richard Fletcher from the Reuters Institute for the Study of Journalism at Oxford University and the University of Copenhagen conducted the research with YouGov administering the survey.

What: A comprehensive survey examining public awareness, usage patterns, trust levels, and perceptions of generative artificial intelligence across different sectors, with particular focus on journalism and news production, revealing 90% awareness of AI tools, 34% weekly usage, and persistent skepticism about AI-generated news content.

When: The survey was conducted between June 5 and July 15, 2025, as a follow-up to 2024 research, with findings published on October 7, 2025.

Where: The research covered six countries representing different media systems and high/upper middle-income contexts: Argentina, Denmark, France, Japan, the United Kingdom, and the United States, with approximately 2,000 respondents in each market.

Why: The research addresses how the public uses generative AI in everyday life, expectations about its impact on different sectors of society, and specific opinions about AI use in journalism, documenting rapid adoption that grew roughly three times faster than initial internet spread while revealing significant comfort gaps between AI-driven and human-led news production.