Generative AI for Information Is Growing, Not Agents

Thirty-eight percent of respondents across countries used generative AI for information in the last week, up from 24% in 2025 and 11% in 2024. That makes generative AI for information a practical product question: people are already using it to research topics, answer factual questions, and get advice, but that does not mean they want AI handling every information task.

Generative AI for information is becoming a routine path

A nine-country report finds conversational AI increasingly acting as an information intermediary. Across countries, 57% of respondents said they saw or engaged with AI-generated answers or summaries in the previous week. That includes encounters through existing services such as search engines and news media, not just direct conversations with an AI tool.

Weekly AI use is also rising: 51% of respondents across the six countries with past data said they now use AI weekly, including 49% in the U.S. Those figures describe different things—general weekly use, exposure to generated answers, and using generative AI to get information—so they should not be collapsed into one adoption number. Together, though, they show that AI is becoming part of how people reach information, both by choice and as a feature inside services they already use.

For someone building an information product, my read is that the useful unit is the task, not the chatbot. Researching a topic or getting a quick factual answer is a clear fit for a lower-friction interface. Whether users will trust the result, return to it, or tolerate errors in messy real-world questions is not answered by these usage figures.

AI agent use and AI-created news remain separate tests

The adoption story changes when the system is expected to act on a user’s behalf or produce news. Across all nine countries, 67% said they had not used an AI agent in the last week. And just 12% were very or somewhat comfortable with news made entirely by AI; that rose to 19% when there was some human oversight. In six core countries, 9% said they used AI for news, up from 6% in 2025 and 3% in 2024.

That is not a contradiction. My interpretation is that people can accept AI as a faster route to information without being ready to delegate actions or accept AI as the author of news. The report also notes a widening gap in shared understanding of LLM capability, and that uneven uptake separates daily users from people who have never used AI. A demo that works for a frequent user may not explain itself—or earn trust—with everyone else.

Build around the information job, not the agent demo

The practical opportunity is to make information tasks easier without assuming users want an autonomous agent or AI-written reporting. I would treat conversational answers and summaries as an interface to evaluate, not as proof that users want a fully automated workflow. The source supports growing use for information; it does not show that people want agents to take over the next step.

For news products, the comfort figures make human oversight a meaningful distinction to test, though they do not establish that oversight alone solves trust. For other information tools, the report gives a reason to test direct question-answering and research flows with both regular AI users and people who have never used it. The signal to watch is whether information use keeps growing while agent use and comfort with AI-created news remain much lower. That gap is where product expectations need to stay grounded.