Google now puts an AI summary at the top of search results — do people actually want to be told it's AI-written?

This explores whether searchers want clear labeling of AI-generated answers in search results, and whether knowing that AI wrote something changes how they treat it. The short version: the collection has no study that asks searchers directly whether they prefer disclosure. It does have research on what happens to attention, trust and learning when AI answers sit at the top of the page. Together that research suggests the label may matter less than you'd expect, at least at first.

The most surprising finding comes from eye-tracking. When AI Overviews appear, they take over the space people look at first. Attention to the top-ranked link drops from 31% to 9%. Yet people rated the AI summary and the ordinary ranked results as equally trustworthy Where do searchers look when AI Overviews appear?. So even though the overview is plainly marked as AI, searchers don't seem to discount it. Click data points the same way. Position-one clickthrough falls about 58% on queries that show an AI Overview How much are AI Overviews actually reducing organic search clicks?, and English Wikipedia lost roughly 5% of its search traffic once the summaries began answering questions on the page Does AI search summaries divert traffic away from Wikipedia?. People see a labeled AI answer and are satisfied enough to stop there.

Work outside search suggests disclosure behaves differently over time. When people are told a partner is an AI, they first avoid it. That bias reverses once they see repeated results showing how well the AI performed. Disclosure alone, without that feedback, didn't help people judge the AI more accurately Does revealing AI identity help or hurt user trust?. Search may be a case where the early skepticism has already worn off. Searchers have seen enough decent overviews that the 'AI' label no longer slows them down, whether or not that trust is earned for a given query.

The less obvious cost doesn't show up in trust ratings. In seven experiments with over 10,000 people, those who learned a topic from an LLM summary reported learning less and felt less ownership of what they knew. The advice they then wrote was sparser than advice from people who used ordinary web search Does learning from AI summaries produce shallower knowledge than web search?. So knowing a summary came from AI doesn't stop it from changing how deeply people engage. Real searchers also don't treat AI and search as either-or. They switch between both Does generative AI chat actually replace traditional search?, and they more often turn to an assistant after searching than before Do people use AI assistants before or after searching?.

The question you may not have known to ask: if labeled AI answers are trusted as much as ranked links, does the label do anything useful? The research hints that a better aid than 'this is AI' would be help judging the answer. One example is 'learning to guide', where the AI points out what's worth examining instead of handing over a verdict, which reduced people's tendency to anchor on the machine's answer Can AI guidance reduce anchoring bias better than AI decisions?. Whether searchers would actually prefer that kind of clarity remains an open question in this collection.