How Google AI Overviews Work (And How to Get Featured In Them)

Introduction: The Answer Is Already at the Top of the Page

Search Google for almost any question today, from “how long does a heat pump last” to “accountant vs bookkeeper for a small business”, and there’s a good chance the first thing you see isn’t a website at all. It’s a written summary, produced by Google’s AI, with a handful of small links to the pages it drew from.

That summary is an AI Overview, and for many businesses it has quietly become the most valuable spot on the page. Google said at I/O 2026 that AI Overviews now reach more than 2.5 billion people every month. If your page is one of the links inside it, you’re part of the answer. If it isn’t, you may be ranking well and still be largely invisible.

The good news is that AI Overviews aren’t a black box. Google has published more about how they work than it has for almost any other search feature, and independent research has filled in a lot of the rest. If you’ve been wondering how to get featured in Google AI Overviews, most of what you need is already in plain sight.

This guide is the second supporting guide in our GEO and AI search series, building on our beginner’s guide to Generative Engine Optimisation (GEO), The Complete Guide to Generative Engine Optimisation (GEO) and our BrightonSEO 2026 recap, How to Get Your Brand Mentioned by ChatGPT, Claude and Google’s AI. We’ll cover:

• What AI Overviews are, and how they differ from AI Mode and featured snippets
• How Google decides when to show one, and how it builds the answer
• Which content signals Google pulls from, according to Google and the latest research
• Why being cited matters, even when clicks are falling
• A practical checklist to improve your chances of being included
• How to control and measure your AI Overview visibility
• The myths worth ignoring

What Are Google AI Overviews?

An AI Overview is a short, AI-written summary that appears at the top of some Google search results, answering the question directly and linking to the web pages that support it. Google describes them as a way to get to the gist of a complicated topic more quickly, with links to explore further.

You may still hear them called SGE. AI Overviews began life as the Search Generative Experience, an experiment first announced at Google I/O 2023. They launched under their current name in the US in May 2024, and reached New Zealand in October 2024 as part of a rollout to more than 100 countries. Google then launched AI Mode in New Zealand in August 2025. So if you’re optimising for “SGE”, you’re really optimising for AI Overviews.

AI Overviews vs AI Mode vs featured snippets

Three Google features get mixed up regularly, and the differences matter because each one picks its sources slightly differently.

Google’s own featured snippets documentation explains how those passages are chosen. AI Overviews work differently: instead of quoting one page, they write a new answer and support it with links. That’s why a page can win a featured snippet and still be missing from the AI Overview for the same search, or the other way round.

The line between AI Overviews and AI Mode is also starting to blur. At I/O 2026, Google announced a redesigned search box designed to make it easier to move from an AI Overview into a follow-up conversation in AI Mode. Google’s guidance treats the two together, and so do we in this guide, but Google is clear that they may use different models and techniques, so the links each one shows can differ.

How Google AI Overviews Work

You don’t need to understand the inner workings of a large language model to improve your chances of appearing in an AI Overview. You do need to understand the basic sequence, because each step suggests something different you can influence. Google set out its own explanation in its guide to optimising for generative AI features, published in 2026, and the steps below follow it closely.

1. Google decides whether an AI Overview is worth showing

Not every search gets one. Google says AI Overviews are only shown when its systems decide they add something beyond the regular results, and that they often don’t appear at all.

The type of question makes a big difference. Seer Interactive’s 2026 study of 53 brands and 5.47 million queries found AI Overviews on around 36% of informational searches, but only around 5% of transactional ones. Question-style searches (what, why, how) triggered one 85.9% of the time, and comparison searches (“X vs Y”) 95.4% of the time. These are Seer’s figures from its clients’ data, so treat them as a guide to the pattern rather than a rule for your industry.

2. It breaks the question into smaller searches

This is the step most businesses don’t know about, and it’s the one that changes the most. Google says both AI Overviews and AI Mode may use a technique called query fan-out: the AI generates a set of related searches and runs them at the same time, to gather more information than the original question alone would find.

Google’s own example is a search for “how to fix a lawn that’s full of weeds”, which might fan out into searches such as “best herbicides for lawns”, “remove weeds without chemicals” and “how to prevent weeds in lawn”. The pages that answer those smaller questions well are all in the running, even if they don’t rank for the original search.

3. It retrieves pages from Google’s own index

The AI doesn’t search the open web on its own terms. Google describes a process it calls retrieval-augmented generation, also known as grounding, which uses its core Search ranking systems to pull relevant, up-to-date pages from its existing Search index. In plain English, the same systems that decide what ranks are deciding what the AI gets to read.

That has one important consequence: if a page isn’t indexed, or isn’t eligible to show in Google Search, it can’t be used in an AI Overview.

4. A Gemini model writes the answer

Google’s Gemini models then review the specific information in those retrieved pages and write a single answer. Google says this grounding step is there to improve the quality, accuracy and freshness of the response, compared with relying on what the model learned in training alone.

5. It links to the pages that support the answer

Finally, Google shows links to the pages that support what the answer says. Because of query fan-out, Google notes that AI features can display a wider and more diverse set of links than a classic web search. That’s the opportunity: you don’t have to be the number one result to be one of the sources.

Paul’s take: “Query fan-out is the bit that changes how we plan content for clients. Google isn’t only matching your page to the exact search somebody typed. It’s looking for the best answers to all the follow-up questions that sit behind it. So the question we now ask is less ‘do we rank for this keyword?’ and more ‘have we properly answered the next three things this customer will want to know?‘”

Which Content Signals Does Google Pull From?

No one outside Google knows exactly how sources are chosen, and anyone claiming to have the full recipe should be treated with caution. What we can do is combine what Google says with what large-scale studies have measured.

What Google says

Google’s position is consistent across its documentation: there are no additional requirements to appear in AI Overviews, and no special optimisation is necessary. The same SEO fundamentals apply, because the AI features are rooted in Google’s core ranking and quality systems.

That said, Google’s 2026 guide is more specific than its earlier advice about what helps. The signals it highlights fall into four groups:

Google’s comparison of commodity and non-commodity content is worth remembering. It contrasts a generic article such as “7 Tips for First-Time Homebuyers” with a first-hand account of a specific decision and what it saved. The first could have been written by anyone. The second could only have been written by someone with real experience.

What the research says

Independent studies add an important twist: ranking in the top ten is no longer the whole story.

• Fewer citations come straight from page one. Ahrefs’ March 2026 analysis of 863,000 keywords and 4 million AI Overview links found that 37.9% of cited pages also ranked in the top ten for the same search, down from around 76% in its July 2025 study. The rest were split almost evenly between positions 11 to 100 and pages outside the top 100 altogether. Ahrefs notes that its methods improved between the two studies, so some of that change may be better detection rather than a pure shift in how Google chooses.
• Other studies point the same way, by different amounts. Search Engine Journal reports that a separate BrightEdge analysis put the top-ten overlap at around 17%, using a different method and dataset. The exact figure varies by study, but the direction is consistent.
• Query fan-out is the likely reason. Ahrefs points to fan-out as one explanation: pages are being cited because they answer one of the related sub-questions well, not because they rank for the original search.
• Video is a meaningful source. In the same Ahrefs data, YouTube accounted for 5.6% of all AI Overview citations, and 18.2% of the cited pages that didn’t rank in the top 100 were YouTube videos.

Our reading of all this is that a strong ranking still helps, because it puts you in the pool Google retrieves from. But it’s no longer the finish line. Covering a topic thoroughly, including the follow-up questions, appears to matter more than holding one position for one phrase.

Why Being Cited Matters, Even When Clicks Are Falling

It’s fair to ask whether AI Overviews are worth chasing at all, given how many clicks they keep for themselves. The data says being cited is far better than being left out, even if it doesn’t bring back the traffic of a few years ago.

• AI Overviews take clicks from the top results. Ahrefs’ updated study of 300,000 keywords found that the presence of an AI Overview correlated with a 58% lower click-through rate for the top-ranking page, using December 2025 data.
• Being cited claws a good share of that back. In Seer Interactive’s 2026 study, brands cited inside an AI Overview earned around 120% more organic clicks per impression on informational searches than brands that appeared on the same page but weren’t cited. That’s roughly 20,700 clicks per million impressions against 9,400.
• It’s still not a return to the old normal. The same study found cited brands still earned around 38% fewer clicks than they would on searches with no AI Overview at all.
• The clicks that do arrive may be better. Google says that when people click through from results pages with AI Overviews, those clicks tend to be higher quality, with people more likely to spend longer on the site. That’s Google’s own claim, so it’s worth checking against your own analytics.

All of these are correlations from large, mixed datasets rather than a forecast for your website. But the direction is clear. When an AI Overview appears, being one of its sources is the best position available on the page, and it keeps your brand in front of the customer even when they don’t click.

If falling clicks are making your reporting harder, our guide to measuring and proving SEO ROI when clicks are declining covers the metrics that tell the fuller story.

How to Get Featured in Google AI Overviews: A Practical Checklist

This is the checklist we work through with clients. None of it guarantees inclusion, because nothing can. What it does is remove the reasons Google might pass you over, and give it good reasons to choose you. Start with eligibility, because nothing else matters if you fail there.

1. Make sure you’re eligible

• Check your key pages are indexed. Use the URL Inspection tool in Google Search Console. A page that isn’t indexed can’t be cited.
• Check nothing is blocking snippets. A page has to be eligible to show with a snippet. A stray nosnippet tag or a very low max-snippet setting can quietly rule it out.
• Check you haven’t opted out. Search Console now has a setting that removes a site from generative AI features altogether (more on that below). If someone switched it on, nothing else on this list will help.
• Check Googlebot can crawl the site. Google specifically mentions making sure crawling is allowed in robots.txt and by any CDN or hosting infrastructure, which is where we most often find accidental blocks.

2. Answer the question, then the next one

• Put a direct answer near the top. If a page answers “how long does a heat pump last?”, say so in the first two or three sentences, then explain.
• Cover the follow-up questions. Think about the fan-out searches behind your topic: costs, comparisons, how to choose, what can go wrong. Answer the ones your customers genuinely ask, on the page or on closely linked pages.
• Don’t create a page for every variation. Google warns that churning out pages for every fan-out query to manipulate AI responses breaches its scaled content abuse policy. One genuinely thorough page beats ten thin ones.

3. Say something only you can say

• Add first-hand experience. Real examples, prices, timeframes, photos from jobs, and lessons from working with customers are what make content non-commodity.
• Back up your claims. Name your sources, include specific figures and explain how you know. Clear, checkable facts are easier for any AI system to use with confidence.
• Keep it current. Update key pages when prices, rules or your services change, and make the update meaningful rather than just changing the date.

4. Make it easy to read and easy to lift

• Use clear, descriptive headings, ideally phrased the way customers ask the question.
• Use tables and lists where they genuinely help, such as comparisons, steps and specifications.
• Keep important information in text. Google asks for important content to be available in textual form, not only inside images, videos or interactive widgets.
• Support the text with good images and video. Google lists them as another way your content can appear, and given how often YouTube is cited, a short video answering a common question can earn its place. Our upcoming guide, The Content Structure That Gets Cited by AI: Answer-First Writing Explained, goes deeper on structure.

5. Get the technical foundations right

• Check key content doesn’t depend on heavy JavaScript to appear. Google can process JavaScript, but Google itself says JavaScript sites are generally more complex to get right for search.
• Provide a good page experience on mobile and desktop. Our Complete Technical SEO Checklist for 2027 covers speed, crawlability and rendering in detail.
• Use structured data that matches the page. Google says no special schema is needed for AI Overviews, but structured data that accurately matches the visible text helps Google understand the page and keeps you eligible for rich results. Our upcoming guide, Structured Data 101: The Schema Types Every Website Needs, explains which types matter.

6. Build your reputation beyond your website

• Keep your Google Business Profile and Merchant Center details up to date. Google names both in its guidance, and AI answers can include local business and product information.
• Earn genuine mentions and reviews. AI features can reflect what’s said about you across blogs, videos and forums. Google warns that chasing inauthentic mentions doesn’t help, so focus on earning real coverage. Our Modern Digital PR Playbook explains how.

7. Track it and keep going

• Check the Generative AI performance report in Search Console monthly to see which pages are appearing (more on this below).
• Search your own priority questions and note whether an AI Overview appears, who is cited and whether what it says about your business is right. Answers change frequently, so look for trends over months rather than reacting to a single day.

Paul’s take: “Whenever a client asks why a competitor is cited and they aren’t, the answer is rarely a technical trick. Nine times out of ten, the competitor’s page simply answers the question more directly and more specifically. When we tighten up the opening of a page, add real numbers and examples, and cover the obvious follow-up questions, that’s usually where we start to see movement.”

How to Measure (and Control) Your AI Overview Visibility

Measuring it in Search Console

For the first time, Google gives site owners an official view of their AI Overview visibility. Google introduced the Generative AI performance report in Search Console in June 2026, and Search Engine Journal reports that Google completed the rollout to all websites worldwide on 31 August 2026.

The report shows how often pages from your site appeared in AI Overviews and AI Mode, broken down by page, country, device and date. There are three limitations worth knowing:

• It shows impressions, not clicks. Clicks from AI Overviews are still counted, but they sit inside the normal Performance report alongside every other web search click.
• It doesn’t show the searches. You can see which pages appeared, but not which queries triggered them.
• Smaller sites may not see it. Google has said sites without enough AI impressions might not get a report.

Even with those gaps, it’s the most reliable starting point available. Comparing which pages appear with which pages bring in enquiries is one of the most useful pieces of analysis you can do right now. It’s also why we’ve added it to the reporting we provide for clients.

Beyond Search Console, the simplest check is still a manual one: search your priority questions regularly and record what you see. Google’s guidance also warns to be wary of third-party tools that claim access to internal Google metrics, because none has it. Tracking tools can be useful for monitoring at scale, but treat their numbers as estimates.

Controlling whether your content appears

Most businesses want more AI visibility, not less. But if you have content you’d rather keep out, such as paid content or information you only want people to read in full on your site, Google offers several controls:

• Preview controls. The nosnippet, data-nosnippet and max-snippet settings limit what Google can show from a page, including in AI features. Using them on a whole page also limits your regular search snippet, so use them carefully.
• Removes a page from Google Search entirely, including AI features.
• The Search Console opt-out. A site-wide setting that excludes your site from AI Overviews, AI Mode and generative AI features in Discover. Google says sites that opt out won’t receive traffic or impressions from these features, and that the choice isn’t used as a ranking signal for the regular results.

One common misunderstanding is worth clearing up. Blocking the Google-Extended user agent in robots.txt does not remove you from AI Overviews. Google says it controls use of your content in some of Google’s other AI systems, while Googlebot remains the crawler for Search, AI features included.

What BrightonSEO 2026 Told Us About AI Overviews

When Paul and I attended BrightonSEO earlier this year, Google’s AI features came up across sessions on content, technical SEO and brand building, and we covered the highlights in What BrightonSEO 2026 Tells Us About the Future of Search and our recap on getting your brand mentioned by AI. Looking back, several points line up closely with what Google has since published:

• Searches are getting longer and more specific. Dr Pete Meyers of Moz made the case that the long tail of search is getting longer, with more conversational, situational queries. That’s exactly the kind of search that triggers an AI Overview and fans out into related questions.
• Accessibility comes first. Janaina Barreto-Romero of Oncrawl showed that content which only appears after JavaScript runs can be invisible to many AI crawlers, and that slow, hard-to-crawl sites are deprioritised. Googlebot handles JavaScript far better than most AI crawlers, but the lesson that your key content should be easy to reach applies to every platform.
• Structured data removes ambiguity. Alex McKenna of Viseon.io and Alex Moss both emphasised schema markup, not because it guarantees a citation, but because it gives machines a clear, unambiguous description of who you are and what you offer.
• Shortcuts don’t work. Thomas Peham of Otterly.ai shared testing that found only 0.1% of AI bot traffic accessed llms.txt files. Google has since confirmed that Search ignores them altogether.
• Reputation is built across the web. Sean Barber of Macmillan Cancer Support summed it up neatly: AI infers credibility from patterns across the web, not necessarily from truth. That’s why consistent reviews, listings and coverage matter as much as your own pages.

Paul’s take: “What struck me reading Google’s AI optimisation guide after BrightonSEO was how closely it matched what the best speakers were saying. No special files, no secret markup, no tricks. Be accessible, be clear, be genuinely useful and be talked about. It’s reassuring for business owners, because it means the work that helps you in AI Overviews is the same work that helps you everywhere else.”

Five AI Overview Myths Worth Ignoring

AI Overviews have attracted plenty of confident advice, and some of it is costing businesses time and money. Helpfully, Google’s 2026 guide includes its own mythbusting section, so on these points we don’t have to guess.

“You need an llms.txt file to appear in AI Overviews.” You don’t. Google says Search doesn’t use llms.txt or any other special AI file, and creating one will neither help nor harm your visibility in Google. It may be worth having for other platforms, which our upcoming guide, llms.txt Explained: Should Your Website Have One?, will cover.

“You need to break your content into small chunks for AI.” Google says there’s no requirement to “chunk” content, and that its systems can understand several topics on one page and show the relevant part. Clear headings help readers, but there’s no ideal page length.

“There’s special schema markup for AI Overviews.” There isn’t. Structured data is still worth using for rich results and clarity, but Google is explicit that no special schema.org markup is needed for its AI features.

“You have to rank number one to be cited.” Ranking well helps, but the Ahrefs data above shows most cited pages don’t rank in the top ten for the exact search. Answering the related questions well can earn a citation from further down.

“You can buy or game your way in.” Paying for inauthentic mentions or mass-producing pages for every possible question doesn’t work in the long run, and the second can breach Google’s spam policies. Anyone guaranteeing AI Overview placement is promising something no one outside Google can deliver.

Where to Go Next in This Series

AI Overviews are one platform in a much bigger picture. The Complete Guide to Generative Engine Optimisation (GEO) sets out the full playbook, including platform-by-platform tactics and a 12-month action plan. If you’re newer to the topic, start with our beginner’s guide to GEO, and Paul’s post on GEO vs SEO explains how the two fit together.

Over the coming months, we’ll be publishing more practical guides in the GEO and AI search series, including:

• ChatGPT vs Google: How Different AI Platforms Choose Which Brands to Mention
• The Content Structure That Gets Cited by AI: Answer-First Writing Explained
• Schema Markup for AI Search: What You Need in 2027
• Building Brand Entities: Why Wikipedia, Wikidata and Knowledge Panels Matter More Than Ever
• How to Audit Your Brand’s Current AI Visibility (Free Framework)
• E-E-A-T in the AI Era: Proving Experience, Expertise, Authority and Trust to Machines
• Local Business AI Visibility: Getting Mentioned in AI Search for ‘Near Me’ Queries

Follow us on LinkedIn so you never miss a post.

Find Out Whether Google’s AI Is Citing You

AI Overviews can feel like Google moving the goalposts again, but the rules underneath are familiar. Be indexed and accessible, answer real questions clearly and specifically, cover the follow-ups, and build a reputation that the rest of the web backs up. The businesses that do this well are the ones Google keeps choosing as its sources.

Book an AI Visibility Audit with Digital Hothouse. We’ll check which of your priority searches show an AI Overview, whether you’re being cited and who is cited instead, review what Search Console’s Generative AI report shows for your site, and give you a clear, prioritised list of changes to improve your chances of being included.

We’d love to work with you. Let’s chat.