The search results page is learning to talk back

Open Google right now and, if you’ve opted into Search Labs, the page looks different. Above the familiar ten blue links sits a generated summary: a paragraph or two of AI-written text, pulled from multiple sources, answering your question before you’ve clicked on anything. Google calls it Search Generative Experience, or SGE, and it has been live in limited testing since Google I/O in May 2023 (Reid). What started as an experiment for a slice of Search Labs users in the U.S. has been expanding steadily since, and it’s worth understanding now, not after it reshapes the traffic every brand depends on.

We’ve spent the last several months watching this shift closely because it touches the work we do every day: building content that ranks, converts, and holds up under scrutiny. SGE isn’t a rumor or a roadmap slide. It’s a real product, already changing what a search results page looks like for the people testing it. The question isn’t whether AI-generated answers belong in search. Google answered that in May 2023. The question is what a brand does about it.

What Search Generative Experience actually is, and isn’t, yet

SGE is not a full replacement for search. It’s an experimental layer, opt-in through Search Labs, that generates a synthesized answer above traditional results for certain queries, then lets users ask follow-up questions in a conversational thread (Reid). Google has been explicit that this is a test, not a launch, and the company has continued to refine it in public. In September 2023, Google rolled out updates addressing how the AI model handles false-premise questions and added an "About this result" feature so users can see how a given answer was generated (Schwartz).

That transparency move matters. It tells us Google knows these answers need scrutiny, and it’s building the scaffolding to provide it before this feature graduates out of Search Labs. We read that as a signal, not a guarantee. Nobody outside Google knows exactly when SGE becomes a permanent fixture in default search results or what it gets called when it does. What we do know is that citation matters inside it. Google’s own February 2023 guidance on AI-generated content made a point of saying rankings reward helpful, reliable, people-first material regardless of how it was produced, not the other way around (Google Search Central). SGE appears to be built on that same premise: reward the source, not the shortcut.

“The question isn’t whether AI”

generated answers belong in search. Google answered that in May 2023. The question is what a brand does about it."

Why click-through rates are about to matter more, not less

Here’s the part that should get a content team’s attention. A January 2024 study reviewed by Search Engine Land found that the links cited inside SGE answers matched a brand’s existing top-10 organic ranking only 4.5% of the time, and matched at the domain level just 1.6% of the time. In other words, 93.8% of the sources SGE cites come from outside a query’s current top-ranking pages entirely (Goodwin). Ranking first in traditional search results is no longer a guarantee of being cited first, or at all, inside the answer that now sits above it.

That’s not a reason to panic. It’s a reason to get specific. The study also found SGE displaying an average of ten links per answer but often only around four unique domains, meaning the same handful of trusted sources get cited repeatedly across similar queries (Goodwin). Earning one of those slots looks less like chasing a keyword and more like building the kind of clearly structured, well-sourced, genuinely useful page that both a person and a language model can extract an answer from quickly. Depth, clarity, and real expertise are doing more work than ever, and vague content built to game a formula is doing less.

How we’re adjusting our approach today

We’re not rebuilding our content strategy from scratch. We’re sharpening it. Three shifts we’ve made across client work heading into 2024:

We write for the question, not just the keyword. SGE answers are built from source material that directly and clearly answers a specific question. We structure content so the answer to the core question is stated plainly, high on the page, in language a system can extract cleanly, then backed with the depth that keeps a human reader on the page.

We treat citations as currency. If a small set of trusted domains gets cited repeatedly, becoming one of those domains is the goal. That means original data, named experts, and claims we can back up with a source, every time. Overviews reward proof. We build proof into everything we ship.

We watch click-through rate as a leading indicator, not a lagging one. We’re tracking it now, for every client, so a shift in SGE’s rollout shows up in the data before it shows up in a quarterly report. Waiting for a trend to become obvious is how brands get caught flat-footed.

“Overviews reward proof. We build proof into everything we ship.”

The work doesn’t change. The delivery does.

Search has changed shape before, and it will change shape again after this. What hasn’t changed is the standard: content that’s specific, well-sourced, and genuinely built to help the person reading it. That was true when the strategy was ten blue links. It’s still true when the strategy includes an AI-generated summary sitting above them. Safe doesn’t survive here, and neither does content built for a search engine instead of a person. We build for both, because the brands that do are the ones still standing when the next shift lands.