AI overview boxes get way more screen time than top search results ever did — but is that just about screen position, not trust?

This explores whether AI content wins attention just because of where it sits on the screen, or whether something else about AI also pulls people in. The most direct evidence here points toward position, though it can't settle the question. In an eye-tracking study of search pages, AI Overviews took over the 'golden' top spot that ranked results used to hold. They received much longer looks, and the share of attention going to the first-ranked result fell from 31% to 9% Where do searchers look when AI Overviews appear?. The most revealing detail is that people trusted AI Overviews and ranked results equally. If trust was the same and attention still shifted, then placement is doing a lot of the work. People seem to look at what's on top, not at what they believe more.

That study shows that people look, not why they keep engaging, and the rest of the corpus suggests position can't be the whole story. People judge an AI they're talking to mostly by how competent it seems (about half of their overall impression), then by how human-like it feels, then by how flexibly it communicates How do users mentally model dialogue agent partners?. Conversational interfaces also tap into skills people have used all their lives for talking to other people. That makes AI feel easy to talk to, even though it doesn't actually communicate the way a person does Why do users fail with AI interfaces designed like conversations?. So the form of the interface, not just its location, invites people in.

There's also a less obvious factor: people sometimes choose machines because machines don't judge them. People who are likely to cheat prefer reporting to an online form rather than to a human, because lying to a machine feels less costly Do dishonest people prefer talking to machines?. That study isn't about AI content specifically. Still, it hints that some engagement with AI comes from what users want to avoid, which has nothing to do with screen layout. Control matters too. People feel more ownership over AI-written text when they had more say in shaping it, while personalizing the model makes no difference Does user control over AI text shape feelings of ownership?. How invested people feel depends on what they get to do, not only on what they're shown.

The surprising part is that the same signals used to measure engagement can be used to shape it. Gaze, hesitation and typing speed can be read as live signals of what someone is thinking. That can help an AI time its help well, or it can be used to profile and steer users Can AI systems read cognitive state from interaction patterns alone?. Eye-tracking findings like the AI Overviews result are therefore more than descriptions. They show where the levers are.

A direct caveat: the corpus has just one study that measures attention by position, and nothing that separates position from novelty, convenience or format in a controlled way. Taken together, the evidence says position is probably the strongest single factor in where eyes go first. Whether people stay engaged likely depends on how competent the AI seems, how it talks, and what it lets them get away with or control.