What it means
Google SGE (Search Generative Experience) was Google's first iteration of AI-generated answers inside Search, launched as a beta in 2023. It produced an AI Snapshot at the top of results, often behind a "generate" button, with a carousel of source links beside it. SGE was the predecessor to AI Overviews and AI Mode, and Kevin spent countless hours stress testing it to work out how the answers were built.
Why it matters
SGE caused a stir because it changed where organic clicks go and forced SEOs to face new problems. Google generates SGE answers by grounding LLMs in search results through retrieval augmented generation, and the sites in the AI Snapshot carousel are not the same links as the classic results. That meant ranking #1 no longer guaranteed visibility in the answer, so the signals that put a site in the carousel became the new question to solve.
Picture testing the keyword "corporate credit card" in the SGE beta. The classic results show the usual brands, but the first card in the AI Snapshot is a site ranking in position 26, which tells you Google is using different signals to assemble the carousel than to rank the page list.
How to use this knowledge
Reverse-engineer the carousel. Compare Google SERP AI output sources against classic rankings to learn which pages Google trusts for the answer, not just the link list.
Invest in reviews. Product, local, and brand reviews gained importance for AI results, so build and surface real review content.
Expect uneven coverage. Plan for new AI-powered SERP results to skip many regulated YMYL queries, which leaves classic organic results intact for finance, health, and legal topics.
Treat it as the on-ramp. Study SGE as the direct predecessor to AI Overviews (AIOs), because the lessons carried straight into AIOs and AI Mode.
Growth Memo guidance
When testing SGE, I learned that the sites shown in the AI Snapshot carousel are not the same links as shown in the classic search results. You would think Google forms AI answers out of the top Search results, but that's not the case.
By not showing AI results by default, Google can test engagement (do users really want AI results?) and save money.
I agree the effect of SGE on the web ecosystem could be devastating. Many publishers, who rely on Google referrals for the majority of their visits, would fold within a few months.
AI Overviews (AIOs) — the productized successor to SGE that replaced the beta.
AI Snapshot — the SGE answer box and source carousel itself.
Retrieval augmented generation (RAG) — the grounding process Google used to build SGE answers.
Gemini — the model that improved SGE quality and speed ahead of its rollout.
YMYL — the sensitive query category where SGE often declined to answer.