If you're searching for "Google SGE," here's the answer in three lines. SGE no longer exists under that name: it was the experiment, and the consumer facing feature is now called AI Overviews. Google has since added a second surface, AI Mode, where search becomes a conversation.
This guide gives you the four dates that matter, what separates the two surfaces, what doesn't separate them (fan out), how to check whether your pages are being cited, and the levers that actually work.
SGE No Longer Exists: What You're Looking For Is Called AI Overviews
The Timeline in Four Dates
May 2023, Google I/O. Google announces the Search Generative Experience (SGE). An opt-in experiment in Search Labs: you had to turn it on to see it.
May 14, 2024, Google I/O. Google rolls out AI Overviews to all US users: "AI Overviews will begin rolling out to everyone in the U.S." The announcement ties the feature back to what preceded it: "People have already used AI Overviews billions of times through our experiment in Search Labs" (blog.google, May 14, 2024). That experiment was SGE. Same product, new name, no more opt-in. In the United States.
March 5, 2025. Google announces AI Mode, a conversational surface presented as "an early experiment in Labs," first opened to Google One AI Premium subscribers (blog.google, March 5, 2025). It only leaves Labs at I/O, in May 2025.
May 2025, Google I/O. AI Mode leaves Labs. After a year among opt-in experiments, Google promotes it to general availability, starting with the United States, where AI Overviews had landed twelve months earlier.
Four dates, two living products. SGE isn't a competitor to AI Overviews: it's its former name.
Why the Word Survives
The term was picked up heavily by SEO articles in 2023 and 2024, and it has stuck around in searches and slide decks. If a vendor is still selling you an "SGE strategy," their vocabulary is two years out of date. The real test lies elsewhere: ask what they've changed in their method since AI Mode left Labs in May 2025. Vocabulary reveals laziness, the date reveals method.
Key takeaway: SGE is the former name for AI Overviews. There was never a separate "SGE" optimization discipline.
AI Overviews: The Summary That Appears Above the Results
A written answer block sits at the top of the results, alongside links to the pages it drew on. The classic blue links don't disappear: they stay, below it.
Two nuances matter. First, the block doesn't trigger on every query, and Google says so: "These overviews only show up when our systems determine that an AI powered response is helpful in answering your query." Second, the trigger isn't stable over time: the same query can show an AI answer one day and not the next, depending on personalization and location. An analysis built on a single screenshot describes one session, not a market.
Where the Cited Links Come From
The surface sits on top of Google's search index. No parallel index, no registration, no feed to submit. Google's documentation sets a two-part condition, and it's the second part people forget: "To be eligible to be shown as a supporting link in AI Overviews or AI Mode, a page must be indexed and eligible to be shown in Google Search with a snippet" (Google, "AI features and your website"). A page that's indexed but has snippets disabled is also out of the running.
A vocabulary point worth being precise about, because it changes what you measure: being cited isn't the same as being mentioned. A model can name your brand without linking a single one of your pages. The indexing requirement applies to the link, not to the mention.
What It Changes for Your Traffic
Organic position one is no longer the first thing the eye meets: it's pushed below an answer that's already been written. On a definition query, that answer can be enough on its own and kill the click. On a comparison or purchase query, the reader wants to see the actual pages and check a price: the click holds up better.
You won't find a CTR loss percentage here, because no reliable, general one exists: trigger arbitration, personalization, and query type move the number too much. The mechanics alone are enough to decide the point: it's your definition pages that are exposed, not your decision pages.
AI Mode: Search Becomes a Conversation
AI Mode isn't a block added above the results. It's a search mode in its own right: the answer takes up the whole page and you can chain follow-up questions. Available on mobile, desktop, and in the Google app since it left Labs in May 2025.
Google puts forward its own figure, which we haven't verified: "On average, a search in AI Mode is three times longer than a traditional search." So these aren't the same questions arriving.
The useful comparison, then, isn't "AI Overviews versus SEO" but AI Mode versus ChatGPT: two conversational interfaces competing for the same moment of use, except one sits on top of Google's index.
Fan Out: The Mechanic the Two Surfaces Share
This is the most common misunderstanding on the topic: query fan-out isn't specific to AI Mode. Google's documentation is explicit: "Both AI Overviews and AI Mode may use a 'query fan out' technique, issuing multiple related searches across subtopics and data sources" (Google, "AI features and your website"). Your question is broken up into sub-queries fired in parallel, and the answer is assembled by cross-referencing their results.
Think through what that implies. Your page can get cited on a sub-question that no one types as such, one you've never tracked a single ranking position for: no keyword tracking report will ever catch that citation. It's the strongest argument for moving from keyword tracking to topic tracking. And it holds for both surfaces, including the one that appears on ordinary queries.
Consequence: a page that answers one narrow intent offers a single point of entry. A page that covers a full topic, broken into self-contained sections, offers as many entry points as it has sections. Fan-out rewards structured coverage.
AI Overviews, AI Mode, ChatGPT, Perplexity: Who Looks for What
- AI Overviews: triggered by a standard query, with unstable Google arbitration. Sources come from Google's index, with fan-out into sub-queries. Your levers: indexing, eligibility for the snippet, structure and substance. Measured in Search Console's generative AI report, impressions only.
- AI Mode: a deliberate switch into the mode, then follow-up questions. Same index, same fan-out. Same levers, with priority given to in-depth topics. Same report, same limits.
- ChatGPT: triggered by a conversation, and the model's decision to go to the web. Sources come from the web, crawled by OAI SearchBot, a display bot distinct from GPTBot. Your levers: OAI SearchBot's access and your presence on the web. Measured through prompt testing and multi-engine scans.
- Perplexity: every question, with answers sourced by default. The web, crawled by PerplexityBot, a display bot not used for training. Same levers, same measurement approach.
What these providers document publicly is their bots. OAI SearchBot "is used to surface websites in search results in ChatGPT's search features," and OpenAI notes that a site blocking it "will not be shown in ChatGPT search answers"; it's a separate bot from GPTBot, which crawls for model training. PerplexityBot, for its part, is "designed to surface and link websites in search results on Perplexity" and "is not used to crawl content for AI foundation models." Both pages tell you how to get into the game. Neither describes how the answer is then put together.
For the two Google surfaces, your lever remains the search index and measurement runs through Search Console. For ChatGPT and Perplexity, it doesn't. That's the real frontier between SEO and the newer discipline of GEO: same goal, visibility; different surfaces, different mechanics.
How to Check Whether You're Being Cited (and Why It Isn't That Simple)
The Manual Test and Its Limits
Type your queries in a private window, note who gets cited. A healthy reflex, but what you get is a snapshot, not a measurement: personalization, location, being signed in, and trigger instability all make the result vary from one session to the next.
What Search Console Shows and Doesn't Show
On June 3, 2026, Google launched Search Generative AI performance reports: a dedicated view of your impressions in AI Overviews, AI Mode, and Discover's AI features, broken down by page, country, device, and date (Google Search Central, June 3, 2026).
Read the metric carefully: impressions, and nothing else. No clicks, no CTR, no queries. Google itself says it's "adding additional metrics over time," so the report is a work in progress. It also isn't available to everyone yet: "We're rolling out this report to a subset of website owners" (Search Console Help). Check first whether it exists in your property. You can see where you appear, not what it earns you or on which query: there's no way to say "this page lost X% because of an AI block."
Looking Across Several Engines at Once
Google isn't your only answer surface. A rigorous approach is to query a bank of predefined prompts, based on your domain and your industry, across several engines such as ChatGPT, ChatGPT Search, Gemini, and Perplexity, and to score each engine's verdict: absent, mentioned, or recommended. What the manual test gives you in one session, a structured scan gives you across surfaces, repeatably.
To set up ongoing measurement, the discipline is the same as for SEO: define the queries that matter to your business, test them on a schedule, and track who gets cited over time. Tooling exists for this; the method matters more than the tool.
What Gets You Cited, and What Does Nothing
The Five Levers That Hold Up
- Being indexable and citable. A page blocked from crawling, not indexed, or with snippets disabled is out of the game regardless of its quality. This is the check to run first: it overrides every other lever.
- Answering within the first lines of the section. The answer first, the context after. A paragraph that opens with "In recent years, the search landscape has..." gets extracted by nobody.
- Breaking content into self-contained sections, titled as questions. Each section should stand on its own outside the article. That's exactly what fan-out goes looking for.
- Providing verifiable factual substance. Sourced figures, procedures, prices, terms, timelines. What gets cited is what states something precisely.
- Existing beyond your own site. A working hypothesis drawn from observation, not a published study: cited pages are rarely those of sites with no trace anywhere else on the web. Press, directories, reviews, syndicated content, and trusted reference platforms. More content on your own domain doesn't substitute for that.
Reddit: The Third-Party Source AI Answers Quote Most
Reddit's threads already dominate the classic results on comparison and opinion queries, and AI surfaces mine them for the same reason: they carry the experience-based judgment that no vendor page can state about itself. A disclosed, consistent presence in the subreddits your buyers actually read is one of the few levers that acts on every answer surface at once. Building it takes community knowledge and patience; specialist partners such as Red Visibility, America's #1 Reddit Agency exist to do exactly that work, and Reddit visibility more broadly.
The False Good Ideas
There's no schema.org markup for "AI Overviews," no registration, no quota to buy. Google's documentation is categorical: "There are no additional technical requirements," and "There's also no special schema.org structured data that you need to add."
Stacking generated pages with no substance of their own: on the accounts we observe, that has never triggered citations. It creates volume, and volume isn't an argument for an engine that has to choose.
Finally, a point few agencies put in writing: Google doesn't publish the selection criteria for cited links. We observe correlations, not the rule. Any promise of guaranteed placement is therefore unverifiable, including ours.
Can You Stop Google From Using Your Content?
The question comes up at every meeting, and the answer changed in June 2026. Three settings coexist, and they're constantly confused with one another.
The Google-Extended token in your robots.txt concerns training, not display. Google's documentation places it under "AI training and grounding in some of Google's other systems," and the Search Console control page confirms it implicitly: "This control doesn't affect AI training; to limit training of the models used to generate responses in Search generative AI features, use Google Extended." Blocking Google-Extended doesn't take you out of AI Overviews. It's the costliest misunderstanding on this topic.
Snippet directives (nosnippet, max-snippet, data-nosnippet) reduce what Google can reuse, but they also apply to the classic snippet. You'd be giving up your showcase in the blue links for a gain on a surface you don't control.
The third setting is new. Since June 3, 2026, Search Console has exposed a "Search generative AI" control that removes your site from AI Overviews, AI Mode, and Discover's AI features. Google states: "This control isn't used as a ranking or inclusion signal affecting other parts of Search." And, the accepted trade-off: "You won't receive any traffic or impressions from these features" (Search Console Help). Like the report, it's rolling out "to a subset of website owners."
The trade-off is no longer "sacrifice your classic snippet or stay in": it becomes an economic one. Situations diverge depending on whether you sell expertise or live off ad-driven audience. Before touching the setting, check its exact state in your own Search Console.
Where to Start
- List the ten queries that drive your revenue. Not the ten that drive volume: the ten that bring in clients.
- For each one, check whether an AI answer appears and who's cited in it. You won't get a measurement (the display varies from one session to the next), but you'll see whether the recurring domains are yours or your competitors'.
- Rewrite the introduction of the matching pages so the answer comes first, ahead of the context and ahead of your company's history.
Start with your ten queries, today. You won't get a measurement, but you'll get the list of domains your customers see you alongside. And that's where everything else begins.
Make Your Narrative Count
Narrative moves markets, and AI surfaces now decide whose version of the story gets repeated. Elyseum works with leaders and companies to build the presence that gets them cited: structured content, independent references, and a narrative that survives extraction by AI engines. Contact our team for an assessment of your AI visibility and a plan to make your pages the ones the engines quote.
Frequently Asked Questions
Yes. SGE (Search Generative Experience) was the name of the opt-in experiment launched in Search Labs, announced at Google I/O in May 2023. On May 14, 2024, Google rolled out AI Overviews to all US users, explicitly tying the feature back to that Labs experiment. SGE is a dead name, not a separate product.
AI Overviews is an answer block displayed above the classic results, which remain below it. AI Mode is a surface in its own right, announced on March 5, 2025 in Labs and then rolled out broadly in the United States at I/O in May 2025. Two surfaces, one shared index, and the same fan-out technique, which Google documents for both.
Since June 3, 2026, Search Console has offered a dedicated report for generative AI features: impressions in AI Overviews, AI Mode, and Discover, broken down by page, country, device, and date. Impressions only (no clicks, no CTR, no queries), and for now rolled out only to a subset of properties. An AI visibility scan, for its part, can query several AI surfaces and return a verdict of absent, mentioned, or recommended for each.
No. No "AI Overviews" schema.org markup, no registration, no feed to provide. Your page needs to be indexed and eligible to display with a snippet, answer early in the section, and bring factual substance. No one knows the exact selection criteria: Google doesn't publish it.
Yes, and since June 2026 without sacrificing your classic snippet. Search Console exposes a control that removes a site from AI Overviews, AI Mode, and Discover's AI features, without acting as a ranking signal outside those surfaces; in exchange, the site no longer receives any traffic or impressions from them, and the rollout remains limited to a subset of properties. This is distinct from the nosnippet, max-snippet, and data-nosnippet directives, which also cut into your classic snippet, and from Google-Extended, which concerns training.