AI information gateway use grows, but audiences still reject AI-made news

Thirty-eight percent of respondents across nine countries said they used generative artificial intelligence to get information in the past week. That is up from 24% in 2025 and 11% in 2024.

The figures come from the Reuters Institute’s Generative AI and News Report 2026, which researcher Felix Simon summarized for Nieman Lab. Simon wrote the report with Rasmus Kleis Nielsen and Richard Fletcher, and the survey included the United States.

Use of standalone tools such as ChatGPT, Claude and Gemini also continues to grow. Across the six countries with historical data, 51% said they use AI weekly, including 49% in the U.S.

People are also encountering AI without deliberately seeking out those tools. Fifty-seven percent said they had seen AI-generated answers or summaries in the past week, with search engines such as Google and Bing serving as key drivers of that exposure.

News products add another layer of AI exposure. On average, 24% of respondents had seen a label saying content was made with AI, while 20% had encountered a bullet-point AI news summary.

AI agents remain less widely used. Sixty-seven percent of respondents said they had not used one in the past week.

When people turn to AI for information, their behavior often resembles traditional search. Twenty-four percent said they use it to research a topic, 18% to answer factual questions and 20% to seek advice.

Using AI directly for news remains less common, although it is growing. Across the six core countries, 9% said they used AI for news, up from 6% in 2025 and 3% in 2024.

People who use AI for news also report relatively high trust in the answers they receive. Among that group, 72% said they trusted the responses.

Reliability, however, was not the main reason people cited for using AI. Only 21% said they used it because it provided more reliable information than other sources, while 64% pointed to the convenience of faster or easier answers.

Nielsen also noted that people who use AI for information report using search engines less. The decline they report for search is larger than the decline they report in news use, suggesting the change extends beyond publishers alone.

Attitudes toward AI-generated journalism have moved much less.

On average, 12% said they were very or somewhat comfortable with news produced entirely by AI. That figure rose to 19% when humans provided some oversight.

The preference for human involvement appeared across every country surveyed. Discomfort was especially high around commentary and opinion.

Simon also highlighted differences in who uses AI. Weekly use is higher among people in professional occupations, while many users rely on free versions of AI tools that may offer fewer capabilities than paid products.

He also cited Andrej Karpathy’s argument that a widening “gap in shared understanding of LLM capability” is emerging between people with different levels of exposure to the technology.

For publishers, Simon argues that AI-generated answers create challenges around discovery, brand attribution, data collection and monetization. Those pressures can increase when users accept an AI answer without visiting the original source.

The findings also highlight a gap between growing AI use and continued skepticism toward AI-produced journalism. Audiences appear more comfortable using AI as a tool for finding information than allowing it to create news without human involvement.

The supplied source does not address whether labeling or disclosure changes how comfortable audiences feel about AI-generated news. It does show that relatively few respondents are comfortable with fully automated journalism.

Simon argues that the divide between frequent AI users and people who never use the technology could have wider consequences. Public institutions may increasingly make decisions about AI that affect people regardless of whether they personally use the tools.