A Reuters Institute survey separates rising AI use from comfort with automated news. Different country groups, different questions—not a worldwide trust scoreboard.
AI-generated conceptual illustration. 51%: weekly standalone AI use across six tracked countries. 12%: comfort with entirely AI-made news across nine countries. Different measures and samples; not worldwide rates.
The adoption department has champagne. The accountability department has a form nobody wants to sign. Before the bolshy percentages start marching across your screen, read the labels.
Two numbers. Two different questions.
The Reuters Institute’s October 29 report puts weekly standalone generative-AI use at 51%, versus 34% in 2025, across six countries tracked in all three annual waves. Daily use reached 28%. Separately, across nine countries, 12% were comfortable with entirely AI-made news; 19% with some human oversight. These are different measures and country groups, not a calculated trust gap.[1]
Across the nine-country sample, 57% reported encountering AI summaries in search during the previous week. Exposure need not mean deliberately opening a chatbot.[1]
Read the small print, droogies
YouGov fielded the online survey August 14–September 28, 2026, with roughly 2,000 respondents per country. Quotas and weights targeted population characteristics, but the panels were non-probability samples and answers included recalled behaviour. A conventional margin of error cannot be calculated. This is neither a worldwide adoption rate nor proof that using AI causes distrust.[1]
The nine countries were Argentina, Australia, Denmark, Finland, France, Japan, Norway, the UK and the US. Australia, Finland and Norway were additions; the trend uses the other six.[1]
Our reading: convenience is not consent
An audience can enjoy a useful machine without handing a newsroom a blank cheque. Clicking a convenient answer is not a vote to dissolve the editor, the correction desk or the responsibility for getting a story wrong.
That is the distinction worth keeping when somebody waves an adoption chart at you as a permission slip. A publication still owes its readers a human address for errors, a visible correction and an explanation of what the machine did. Distribution is a technical achievement. Earning confidence is a different job.
FOLLOW THE SOURCES.
Checked Oct 29, 2026. Numbered references point to the evidence behind factual passages. Commentary is our own.
Developers can call Alibaba’s new flagship through an API. It scored slightly above the median for similarly priced models in one independent test, where its output token use was also above the median.
A Fed report puts advanced computing debt issuance above $680 billion by October. It also spells out how leases and guarantees can tie tech companies to projects they did not finance directly.