Reuters' Generative AI and News Report 2025 (reutersinstitute.politics.ox.ac.uk)

Reuters’ “Generative AI and News Report 2025” surveys six countries and finds rapid normalization of consumer generative AI: awareness rose from 78% to 90%, ever-use jumped from 40% to 61%, and weekly use nearly doubled (18%→34%). ChatGPT remains the dominant active brand (22% weekly), while information-seeking is now the leading use-case (weekly 24%), overtaking media creation. AI-generated search answers are widespread (54% saw one in the last week) but engagement and trust are mixed—only ~33% say they often click source links, ~50% trust AI answers, and verification spikes in high‑stakes domains. Use of AI for news doubled but is still niche (3%→6% weekly), and a strong “comfort gap” persists: 62% prefer fully human-produced news versus 12% comfortable with fully AI-generated content; comfort rises markedly when humans lead or oversee AI. Back-end tasks (grammar, translation) see far higher acceptance than front-facing roles (artificial presenters). For the AI/ML community these findings stress practical priorities: invest in provenance, source citation, and UI nudges to encourage click-through and verification; design human-in-the-loop workflows and transparent labels to close the comfort gap; and tailor models and deployment strategies for demographic and domain variance (younger users adopt faster; trust collapses in politics/health). Product metrics should weigh not only accuracy but explainability, verifiable citations, and user trust signals—especially for news and other high‑stakes applications.