SGE: The Announcement That Made Every SEO Update Their CV

Written in 2026, looking back at the Search Generative Experience reveal, now that I’ve lived through what it became.

Google I/O on May 10, 2023 was the first time the company showed AI answers sitting inside the search results themselves. They called it the Search Generative Experience, SGE, which is a name only a committee could love. A coloured box at the top of the page, generating a paragraph of answer, with the blue links pushed down beneath it. I watched the stream from Bali, and I could almost hear a whole industry opening a browser tab to freshen up their CV.

I want to be honest about what I thought would happen, because it’s easy to look prophetic in hindsight and I wasn’t, fully.

What I predicted in May 2023

Three things, roughly, and I’ll score myself later.

First, I thought this would gut informational traffic. If the answer to how long to boil an egg sits in a box at the top, nobody scrolls to the article about boiling eggs. That felt obvious and I said it with confidence.

Second, I thought it would ship slowly and stay ugly for a long time. Google had a business built entirely on people clicking links, and this box was, structurally, a machine for stopping the click. I could not see how they’d square that with the revenue, so I guessed they’d hesitate, water it down, keep it in a limited test for ages.

Third, I thought the winners would be brands people already trusted, because a generated answer needs sources it can stand behind, and Google would reach for names it recognised. I told clients the safest place to be was known.

What actually shipped

I was right about the first thing and it was worse than I said. The informational drop, once these answers rolled out properly, was brutal in the categories they touched. The click and the ranking became two separate things you could win or lose independently, which I’ve written about since because it reshaped how I even report results.

I was half right about the second thing. Google did hesitate, kept SGE as an opt-in experiment for a long stretch, exactly as I guessed. But I underrated their nerve. They didn’t quietly kill it to protect the click, they leaned in, and by the following year the AI answers were a default fixture rather than a lab toy. My instinct that the revenue would save the links was wrong. The fear I keep noticing in Google since the Bard demo won out over the caution.

Third thing, mostly right, and it’s the prediction I’m least embarrassed about. Trusted, known, first-hand sources did become the safer place to stand. The sites that got cited by the AI answer tended to be the ones with a real reason to be believed.

The lesson I actually kept

The useful thing wasn’t any single prediction. It was noticing, in May 2023, that Google had decided the answer mattered more than the destination. Once a company decides that, everything downstream follows, and you can plan around it even when your specific forecasts miss.

So the CV panic in my industry was understandable but aimed at the wrong worry. The job that was ending was ranking a page to be visited. The job that was starting was being the source an answer is built from. Same skills, mostly, pointed at a new target. I said some version of that at the time, a bit too quietly, hedged with too many maybes. I’d say it louder now.