How to Rank in Google AI Overviews

Google AI Overviews are the AI-generated summaries that sometimes sit at the very top of a Google results page, above the traditional links, answering a query directly and citing a handful of sources. If you know the feature as SGE, this is the same thing: Search Generative Experience was the name of the experiment, and AI Overviews is the product it became. This guide is for marketers, founders, and content leads who want an honest read on how those summaries are built, what content gets pulled into them, and what you can actually influence. It covers how an Overview is generated, why ordinary ranking still does most of the work, the formatting that makes a passage easy to lift, the trust signals that matter, and the zero-click trade-off to weigh before treating Overviews as a goal. Where something is confirmed by Google, this piece says so; where it is an observation from working with search, it says that too.

The short version

• SGE and AI Overviews are the same feature: SGE was the 2023 experiment name, AI Overviews is the product it shipped as. Either way there is no separate button to press. Google states that AI features run on the same ranking and eligibility systems as the rest of Search, so appearing in an Overview starts with ranking well for the query.
• An Overview is retrieval plus generation: Google pulls candidate passages from its index, then a model synthesizes them into a summary with links. Being in the index and near the top is the price of entry.
• Clear, self-contained answers get pulled. A passage that resolves one question without needing the rest of the page is the unit an Overview reaches for.
• Schema and formatting help a machine parse and lift your content. Neither buys a citation, and Google does not require special markup to appear.
• Appearing is not always a win. For some queries an Overview answers the question so completely that the click you used to earn no longer happens.

What exactly is a Google AI Overview?

It is an AI-generated answer block that appears above the standard results for some searches, summarizing information drawn from multiple web pages and linking to them as sources. Google introduced it as part of bringing generative answers into Search, and it does not show for every query. Informational and how-to questions tend to trigger it; many navigational, transactional, and highly specific queries do not.

The important framing: an Overview is a new way to present answers, not a new place you submit content to. The pages it draws from are ordinary indexed web pages competing in ordinary search. The same logic extends to spoken answers: voice assistants pull from this same retrieval layer rather than running a separate voice-specific ranking system. Our broader take on this shift lives in our comparison of SEO and AI search, which covers why the two run on one shared index rather than against each other.

Is SGE the same as AI Overviews?

Yes. SGE, short for Search Generative Experience, was the name Google used while the feature was an opt-in experiment. AI Overviews is the name it shipped under. Google announced SGE in May 2023 as a Search Labs experiment users had to switch on themselves, then began rolling AI Overviews out to everyone in the United States in May 2024. The experiment name retired with the experiment.

The distinction matters for one practical reason. Advice written for SGE in 2023 describes an opt-in lab product that a small pool of users had switched on, not the default results page millions of people now see. Treat those guides as historical. If you are searching for how to rank in SGE, you are asking how to rank in AI Overviews: earn a strong ordinary ranking for the query first, then structure the page so a clean, self-contained answer can be lifted from it. The rest of this article covers both halves.

Neither name should be confused with AI Mode, which is a different surface rather than another rename. An AI Overview appears inside the ordinary results page, above the links. AI Mode is a separate tab built for conversational, multi-step questions, and it replaces the results page rather than sitting on top of it. Google now documents both surfaces in the same guidance for site owners, and the advice it gives for each is the same, which is the useful part. The same naming confusion runs through the wider vocabulary around AI search, and our guide to what the difference between AEO and GEO actually changes covers which of those labels describes work on your own page and which describes work off it.

How does Google generate an AI Overview?

In two stages: retrieval, then generation. Google first retrieves candidate pages and passages from its existing index, the same index that powers the blue links, often by running several related searches behind a single prompt rather than one, a mechanism our guide to query fan-out covers in full. Then a model selects the parts it judges most relevant and trustworthy and rewrites them into a single summary with citations.

This matters for one practical reason. The retrieval stage is bounded by what already ranks. Google's guidance on AI features and your website is direct about this: ordinary SEO best practices still apply, a page has to be indexed and eligible to be shown with a snippet, and there are no additional requirements to appear in AI Overviews or AI Mode. A page that is not indexed, or that no ranking system surfaces for the query, is not a candidate the generation stage can even see. You are optimizing the input, not the output.

Do you need to rank well to appear in an AI Overview?

Generally, yes. Because retrieval draws from pages that already perform for the query, the pages cited in Overviews are usually pages that also rank on the first page of ordinary results. This is an observed pattern rather than a published rule, but it is consistent enough to plan around: earning a strong ordinary ranking is the most reliable path into the Overview for that query.

The corollary is freeing. You do not need a separate AI Overviews strategy sitting apart from your SEO; you need content that ranks, structured so a model can lift a clean answer from it. If a page cannot rank in classic search, optimizing it for Overviews is optimizing something nothing will retrieve. Our AI SEO guide covers the answer-engine side in more depth, and our SEO services page outlines the ranking foundation it sits on.

What kind of content gets pulled into an AI Overview?

Content that answers one question cleanly, in a passage that stands on its own. A model building a summary is looking for a block it can quote or paraphrase without distortion, so the more self-contained and specific a passage is, the more usable it becomes. Our guide to answer engine optimization turns that requirement into a six-check test you can run on a passage before you publish it. That requirement rules out audio and video by default: a podcast episode with no transcript gives a retrieval system nothing to read, which is why treating the transcript as the primary asset matters as much for AI answer engines as it does for Google web search. The framework below maps the signals we see in cited content against what they look like on the page.

Signal an Overview reaches for	What it looks like on the page	Why it gets pulled
A direct answer	The first sentence under a heading answers the heading, before any warm-up.	The model can lift it whole and trust it resolves the query.
A self-contained passage	A 40 to 90 word block that makes sense with no surrounding context.	Retrieval works on chunks, so a passage is only as usable as it is standalone.
Topical depth	The page covers the question and its related sub-questions and entities.	Depth signals the page is genuinely about the topic, not brushing past it.
Credible sourcing	Specific, attributable claims, with facts tied to their source in the sentence.	A checkable claim adds something to the answer that a vague one cannot.
Structured data points	Comparisons in tables, steps in lists, thresholds and versions named explicitly.	Models parse and reproduce structured facts more reliably than prose.

Notice what is not on the list: keyword density, word count for its own sake, and clever markup. Those are inputs people over-index on because they are easy to measure, not because they decide what gets quoted.

How much do headings, lists, tables, and formatting matter?

They matter for extraction, not for selection. Formatting does not persuade Google to choose your page, but once your page is a candidate, clean structure decides whether a usable passage can be lifted from it. This is the same instinct that wins featured snippets, and the overlap is not a coincidence: both reward a direct answer in a clearly bounded block.

The formatting that consistently helps a passage travel:

• Question-shaped headings: Phrase an h2 the way a person asks the question, then answer it in the first sentence.
• Short answer blocks: Keep the core answer to roughly 40 to 90 words so it can be quoted whole.
• Tables for comparisons: A three-column comparison is easy for a model to read and reproduce accurately.
• Lists for steps and criteria: Ordered steps and clear criteria parse cleanly.
• Literal terms named exactly: Use code for the precise property, threshold, or file rather than describing it loosely.

Schema markup belongs in this bucket too, with an honest caveat.

Does schema markup help you appear in AI Overviews?

It helps a machine understand your page; it does not buy a place in the Overview. Google's structured data documentation describes markup as a way to make content eligible for rich results and to help Search understand what a page is about. It is not documented as a lever for AI Overview inclusion. Google's AI features guidance goes further and says outright that there are no new machine readable files, AI text files, or markup to create in order to appear in AI Overviews or AI Mode.

So treat schema as plumbing, not persuasion. FAQPage where real questions appear, Article with a real author and dateModified, and Organization sitewide all help machines parse and attribute your content correctly. That supports clean extraction. It does not substitute for being the best answer. Our schema markup guide covers which types actually earn their keep. If you are wondering about the llms.txt proposal, treat it as optional: no major engine has committed to it, so it is worth knowing about but not worth prioritizing.

What role do entity and E-E-A-T signals play?

They influence whether Google trusts your page enough to build an answer from it. Google's Search Essentials still set the baseline for being eligible to appear anywhere in Search, and the experience, expertise, authoritativeness, and trust framework still shapes which sources its systems reach for. AI Overviews did not retire any of it.

Two things carry weight here. First, entity consistency: naming your brand, your people, and your services the same way across your site, your schema, and your off-site profiles helps Google connect the page to a recognized entity. Second, demonstrated expertise on the page itself: a real author, specific claims, and evidence a competent reader could check. A page that reads as anonymous and generic is a weaker candidate than one clearly written by someone who knows the subject, even when they cover the same facts.

How should you think about the zero-click reality?

Accept that an Overview can answer the question without sending a click, and change what you measure accordingly. When a summary resolves a query outright, the informational visit you used to earn may not happen, and that is most pronounced on top-of-funnel "what is" and definitional queries. Pretending otherwise leads to disappointment.

The more useful frame is brand visibility and qualified clicks. Being named as a source builds awareness and trust even without a visit, closer to earned media than to a ranking. The clicks you do get tend to come from people who read the summary and still wanted more, which is often a more qualified visit than a casual scan of ten links. The trade is fewer shallow visits for more brand exposure and better-qualified ones, and whether it is a good trade depends on your funnel.

The AI Overviews readiness checklist

Work through this on a page you already rank with. It will not manufacture a ranking you have not earned, but it makes an already-competitive page far easier to pull from.

Foundation: The page is crawlable and indexed. It already ranks on page one for the target query, or is close. The main content is in the server-rendered HTML, not injected after load, because several crawlers render JavaScript poorly. You are not blocking Google crawlers in robots.txt.

Extractability: Every h2 is a real question. The first sentence under each one answers it completely. Each section stands alone, so a reader arriving at the seventh heading needs nothing from the first. Core answers sit in roughly 40 to 90 word blocks.

Substance: Facts are specific and attributed in the sentence. The page covers the main question plus its obvious follow-ups. At least one comparison table and one original element (a framework, a checklist, a dataset) give the page something quotable that exists only here.

Trust and parsing: A real author with relevant expertise is named. Article, FAQPage, and Organization schema are present and accurate. Your brand and people are named consistently across the site and off-site profiles. Content is fresh, with an honest dateModified.

If a page fails the foundation group, stop. No amount of formatting compensates for a page that does not rank.

How do you measure whether you appear in AI Overviews?

Imperfectly, and it is worth admitting that up front. Google folds clicks and impressions from AI features into the overall Web search type in the Search Console performance report, with no separate breakout, so no report cleanly attributes an AI Overview appearance. Measurement leans on a mix of methods, none of them precise. Anyone quoting you an exact "AI Overview share" is estimating.

Method	What it tells you	What it cannot tell you
Manual query checks	Whether an Overview shows for your target queries, and whether you are cited.	Overviews vary by user, location, and over time, so one check is a snapshot, not a rate.
Search Console trends	Impression and click shifts on informational queries where Overviews are common.	AI feature traffic sits inside the same Web search type, so it cannot be separated out.
Brand-search lift	Whether branded searches rise as your unlinked visibility grows.	It is a lagging, indirect signal, easily confused with other marketing.

Track direction over months, not a figure on any given day. Record your spot checks so a "we appeared" or "we lost the citation" is evidence, not memory. Do not invent a percentage to report upward; a labeled spot check is more honest and more useful than a fabricated metric.

When should you not chase AI Overviews?

When appearing would cannibalize the clicks that pay your bills. For a page whose value is the visit itself, a purely informational post that monetizes through on-page conversion, winning the Overview can mean the answer is consumed and the visit never happens. In that case, optimizing hard for extraction can work against you.

Hold off, or at least think twice, when any of these is true:

• The query is transactional or commercial, and the click is the whole point. Overviews are less common here anyway, and a lost click costs more than a citation is worth.
• The page cannot yet rank in ordinary search. Fix the foundation first; there is nothing for an Overview to retrieve.
• Your audience does not use AI-assisted search for this topic. Some local and niche commercial queries still happen almost entirely on classic results.
• Your model depends on ad impressions from informational traffic. Weigh the brand exposure against the impressions you would forgo.

This is a genuine trade-off, not a tactic to switch on everywhere. Decide it per page, against what that page is for.

What can you not control about AI Overviews?

More than you can, and naming it keeps expectations honest. You cannot control whether Google shows an Overview for a query, which sources it selects, how it phrases your content, or whether it links to you when it uses your information. These are Google's systems, and they change without notice.

What you do control is the input: whether your page is crawlable, indexed, ranking, trustworthy, and structured so a clean answer can be lifted from it. That is the whole of your leverage, and it is the same leverage that has always driven organic visibility. The surface changed. The work did not.

The one habit worth building: treat AI Overview readiness as a byproduct of ranking well, never as a separate campaign, and revisit it only on pages that already earn their place in results. This week, take your three highest-value informational pages, read the first sentence under every heading, and rewrite any that do not answer the heading on their own. Then decide, page by page, whether being quoted actually serves that page's goal, or whether the click is worth protecting.