Generative search has been telegraphing itself for a year
Markup refreshed for current answer engines.
It is easy to treat AI in search as a someday problem, because the version most people can see is still labelled an experiment. But Google has been testing a generated answer above the links — SGE — in public for nearly a year. The launch date is uncertain; the direction is not.
the short answer
Since May 2023, Google has tested SGE (Search Generative Experience) in Search Labs: a synthesized answer with cited sources, above the blue links, for informational questions. As of April 2024 it is still experimental, but Google removed the experiment’s end date in December 2023. The direction — answer first, sources cited — is clear a year ahead, and preparing to be a source it can quote, while it is still experimental, is an advantage.
key takeaways
- Generative search did not appear from nowhere: since May 2023, Google has been testing it in public as SGE (Search Generative Experience) inside Search Labs.
- SGE puts a synthesized AI answer, with cited sources, above the traditional blue links — mostly for informational questions — while the links still appear beneath.
- As of April 2024 it is still experimental (opt-in, 120+ countries excluding Europe), but Google removed the experiment’s end date back in December 2023 — a signal it is not a throwaway.
- The direction is legible a year ahead: from "search and choose among links" to "receive an answer with sources." The question is not whether, but when.
- Whoever prepares while it is still experimental — clear content that answers questions, credible sources — starts ahead of whoever waits for the launch.
a year of public signal
Read left to right, the signal is hard to miss: a public experiment that was meant to end in 2023, kept alive and refined into 2024 across a large test surface. The exact launch date is not on this timeline, because it is not yet known — but the arrow’s direction has been steady for a year.
Why timing uncertainty is not a reason to wait
The reasonable objection to all this is that nobody knows when, or even whether, a generated answer becomes the default search experience — and that is fair. The timing is genuinely uncertain; people who follow this closely disagree about it, and anyone claiming a firm date is guessing. But uncertainty about timing is often treated as permission to do nothing, and that is the error. The cost of preparing early for a shift that is clearly coming is low, because the preparation is useful on its own terms: clearer content, answered questions, credible sourcing all serve the audience and the search you have today. The cost of waiting until the shift is undeniable is high, because by then everyone moves at once, into a surface where standing as a trusted source is earned slowly. When the downside of acting early is small and the downside of acting late is large, uncertainty about the exact date argues for moving now, not for holding back. The asymmetry is the whole argument: you are not betting on a date, you are buying cheap insurance against a direction that is no longer in serious doubt.
It also reframes what you are preparing for. The instinct is to ask "how do I optimise for SGE," as if it were a specific feature to game, but the feature will change names and mechanics before it settles — what is stable is the underlying move from listing documents to synthesizing an answer that cites sources. Preparing for that is not about chasing one product’s quirks; it is about becoming the kind of clear, credible, quotable source that any answer-first system, from any vendor, has reason to draw on. That framing protects you from wasting effort on the specifics of a still-changing experiment, and points it instead at the durable thing underneath. SGE is the version visible today; the direction it points to is bigger than SGE, and that direction is what rewards early, patient work.
The read, in three parts
What SGE is and what it changes, how to read the signal it sends, and how to prepare while it is still experimental. Open each layer for the part that changes how you weigh the timing.
01 What SGE is, and what it changes
SGE — Search Generative Experience — is Google’s experiment in putting a generated answer at the top of the results page. When you ask an informational question, instead of going straight to ten blue links, you get a synthesized summary assembled from several sources, with those sources cited, and the familiar links sitting below it. That is a different shape of search. The old model handed you a set of documents and left the synthesis to you: open a few, read, decide. The new one does the synthesis first and hands you an answer, with the documents demoted to support and citation. For a reader, that is often faster. For anyone who depends on being found, it is a shift in where attention lands — from "do I rank among the links" to "am I one of the sources the answer is built from and cites." The mechanics are still settling and the feature is still labelled an experiment, but the core move — answer first, links second — is already visible in how SGE works today. And once you have seen that move, you cannot un-see it: every refinement Google ships to the experiment is a refinement to answer-first search, not a step back toward the ten-blue-links page it is quietly leaving behind. The experiment label is a description of how widely it is switched on, not a hedge about whether the underlying idea is going anywhere; the idea has been steady even as the rollout has been cautious.
02 Reading the signal
The temptation is to wait until something is officially launched before taking it seriously, and with SGE that instinct misleads, because the public record already tells you plenty. Google introduced it nearly a year ago and has been refining it in the open ever since, across a large, multi-country test surface. It was first framed as an experiment scheduled to wrap up at the end of 2023 — and then, in December, Google quietly removed that end date. A company winding an experiment down does not extend it indefinitely and keep investing in it; a company that intends to ship does. None of this tells you the exact launch date, and honest analysis should not pretend otherwise: the timing remains genuinely uncertain, and reasonable people in the field disagree about it. But timing and direction are different questions. The direction has been telegraphed clearly and repeatedly for a year — search is moving toward answer-first, with sources cited — and you do not need the launch date to act on the direction. Treating the missing date as a reason to ignore the visible direction is the kind of mistake that looks prudent right up until the moment it looks slow.
03 Preparing before the launch
Acting on the direction, while it is still experimental, is the whole advantage. The work is not exotic: it is becoming the kind of source a system can quote with confidence, which is also the kind of source a human is glad to find. Lead with a clear answer to the real question instead of burying it; cover your topics with genuine depth and accuracy; carry the signals of a credible source rather than reading as thin, generic filler. Make sure that, for the questions your audience actually asks, your content answers them better and more plainly than the alternatives — because in an answer-first world, the gap between "adequate" and "clearest" is the gap between being cited and being skipped. The reason to do this now, rather than when a launch makes it urgent, is that none of it is wasted in the meantime: the same qualities that earn a citation in a generated answer serve the readers and the search you already have. Early preparation is cheap and it compounds; the established, trusted sources are built over time, not conjured the week a feature goes mainstream, which is precisely why the brands that move first tend to hold the advantage long after the rest catch on. This is the kind of forward read the AC Group has made its work for 27 years — seeing where attention is moving and helping clients be ready when it arrives.
What "preparing early" actually looks like
Preparing early can sound abstract, so it is worth making it concrete, because the work is specific and most of it you can start this quarter. Begin with the questions: list the things people genuinely ask in your category — the comparisons, the how-tos, the "is X right for Y" decisions — and check, honestly, whether your content answers each one clearly and near the top of the page, or whether the answer is buried under preamble. A generative system rewards pages that lead with the answer, so the simple act of moving your clearest sentences up and stating them plainly is among the highest-return changes you can make. Then look at depth and credibility: is each topic covered with enough substance and accuracy that a system would have reason to quote you over a thinner competitor, and do your pages carry the marks of a source worth trusting rather than generic filler?
From there it becomes a short, repeatable loop rather than a one-off project. Try the questions in the generative surfaces you can already reach, note where you are cited and where a competitor is named instead, and treat each absence as a specific gap to close with clearer, deeper, more credible content on that exact question. Because the surface is still experimental and changing, do this periodically rather than once — the point is to build the habit of being a clear, quotable source, not to chase a single configuration that will be different next quarter. The compounding is the reward: every gap you close is durable, serving the search you have today and the answer-first search that is coming, and the sources that are trusted when the shift becomes mainstream are the ones that started earning that trust while everyone else was still waiting for a launch date.
SGE and the direction of search: quick answers
Is SGE live for everyone right now?
No — as of April 2024 it is still an experiment. SGE lives inside Search Labs, Google’s opt-in program for trying early-stage search features, so you only see it if you have signed up and you are in a supported region. It has been available across more than a hundred countries and territories, notably excluding Europe, and runs on Chrome desktop and the Google app. That experimental status is exactly why it is easy to dismiss: if it is not in front of most users yet, it can feel like a curiosity rather than a plan. But "experimental" describes its distribution, not its seriousness. Google has been refining it in public for nearly a year, and the testing surface is large and multi-country, which is not how a company treats something it intends to shelve. The honest reading is that it is not live for everyone, and that this is a window — the time to understand it before it becomes the default, rather than after.
Will this replace the blue links?
Not replace, at least not in the form it currently takes — supplement is the better word, and the distinction matters for how you respond. In SGE today, the generated answer sits at the top as a synthesized summary with cited sources, and the familiar list of links still appears beneath it. So the model is additive: a new layer above the results rather than a deletion of them. What changes is attention and order. When a usable answer appears first, with sources named, a share of users get what they need without scrolling into the links — which redistributes visibility toward whoever is cited in the summary and away from whoever ranks tenth in a list fewer people reach. So the right question is not "will links disappear" but "what happens to my visibility when an answer comes first." Planning around that shift is more useful than betting on whether the links survive in their current form.
If it is still experimental, should I do anything now?
Yes, and the experimental phase is the advantage, not a reason to wait. Two things are true at once: the timing of any broad launch is genuinely uncertain, and the direction is not. When the destination is clear but the date is not, the move is to prepare in a way that pays off regardless of exactly when the switch flips — which here means becoming the kind of clear, credible source a generative system can quote, work that also serves the search you already have. There is no penalty for being ready early; the content that earns a citation in a synthesized answer is the same content that serves a reader well today. The risk runs the other way. If you wait for the launch to be undeniable, you start the work at the moment everyone else does, into a surface where being an established, trusted source takes time to build. Early is cheap and compounding; late is crowded and slow.
How do I become a source SGE cites?
The dependable answer is unglamorous: be the clearest, most credible source on the questions your audience actually asks. A system that synthesizes an answer and names its sources is looking for material it can quote with confidence — content that states things plainly, answers the real question directly, and comes from somewhere it has reason to trust. That favours pages that lead with a clear answer rather than burying it, that cover a topic with genuine depth and accuracy, and that carry the signals of a credible source rather than thin, generic filler. None of this is a trick specific to SGE; it is the same quality that has always made content worth surfacing, now read by a system that quotes rather than just ranks. The practical starting point is to look at the questions in your category, ask whether your content answers them better and more clearly than the alternatives, and fix the places where it does not — because that gap is exactly where the citation goes to someone else.
A note on sources and timing
This is written in April 2024, and it describes the state of things as they stand now. SGE has been a Search Labs experiment since May 2023; it presents a generated answer with cited sources above the traditional links, mainly for informational queries; it remains opt-in and unavailable in Europe; and Google removed its scheduled end date in December 2023. Microsoft has been pursuing the same direction from its side, integrating OpenAI’s models into Bing. We have not asserted a launch date or a rebrand for SGE, because as of this writing neither has happened and predicting them would be guesswork dressed as fact — what we can say is that the direction has been signalled clearly for a year. We have also kept the advice to the durable part — be a clear, credible, quotable source — rather than to the specifics of an experiment that will keep changing. Reading where attention is heading and helping clients be ready before it arrives is work the AC Group has done for 27 years, across more shifts in how people find information than this one, and the next reading of the data will only confirm what the early direction was already showing.