Our research of 405,576 searches showed that Google AI Overviews (AIOs) appear in roughly 47% of all Google searches, with the number being as high as 70% for informational queries. It's clear that Google is betting on AI, so you need to go beyond traditional search results and optimize for AI-powered engines, as well.
Seeing as AIOs have been around for some time now, we have an idea of how to get cited or featured in them. In this guide, I'll share everything we know to explain what actually drives AI Overview citations based on data and research, not speculation. I'll then share a few steps you can take to boost your chances of getting cited and improve your AI Overviews rank.
How does Google select sources for AI overviews?
Google uses a process called retrieval-augmented generation (RAG) to create AIOs for the given queries. Here's a quick run-through of how it works:
- The algorithm retrieves candidate pages and ranks them semantically
- Large language models (LLMs) organize and prepare the response
- Google's generative AI writes the response and cites its sources
You'll see all the cited sources on the right side of the response, like so:
And if you scroll through the SERP, you'll see that organic rankings are often much different from the sources.
This is because AIOs don't care about the exact query/keyword or use it as the basis of their research. Instead, they dive deeper using a process called a query fan-out.
How does query fan-out work?
Query fan-out breaks a single query into multiple sub-queries to cover as much related information as possible. It helps search engines understand the query on a deeper level and gather enough context for a comprehensive response.
For example, if I Google "What is the best CRM for small businesses," the algorithm might research stuff like:
- The pricing of popular CRM platforms
- Specific features tailored to small businesses
- Key considerations when choosing a CRM
Our query fan-out research showed that most queries trigger up to around five fan-outs. More importantly, we saw that a page's chances of being cited increase by 161% if it ranks for at least one of those fan-outs on top of the main query.
While I'll talk about this more when I share some tips on ranking in AIOs, here's a bit of a spoiler we can take away from the study: if you research and target fan-outs, you're much more likely to end up in AIOs' sources.
Where do citations actually come from?
For better or worse, citations don't necessarily come from SEO-friendly sites. Our AI citations study showed that YouTube (~23.3%), Wikipedia (~18.4%), and Google.com (~16.4%) account for around 58.1% of all AIO sources.
AIOs also rely heavily on forums and social media platforms, most notably:
- Quora
While it may seem like this doesn't leave much room for actual web pages to be cited, that's not really true. Citations often go to specific domains, especially those with deep expertise related to the query. All it takes is to show Google that your pages are among them.
1. Use query fan-outs for deep topical authority
Topical authority showed up as a concept way before AI, but it's more relevant than ever. Due to query fan-out, sites with comprehensive coverage of a topic are much more likely to rank for multiple queries and get cited by AIOs.
Our data from the study I mentioned clearly confirms this. Out of all the AIO citations that rank in traditional search results, 51.2% rank for the main query and at least one sub-query. Meanwhile, only 19.6% rank for just the main query.
The best way to cast a wide net with your content is through topic clusters. Instead of a huge mega-guide covering every single aspect of a topic, split your content into the main page and the related cluster pages.
For example, if your main page is about cloud storage solutions, it can cover some general information but be supported by dedicated cluster pages, each covering a topic like:
- Types of storage
- Cost comparisons
- Security best practices
- Migration tips
With a solid cluster, AI systems keep stumbling upon your content as they fan out the query and look for specific answers. This drastically increases your chances of having at least one of those pages cited.
You can use Surfer's free Chrome keyword extension to find query fan-outs in AI answers. Just open ChatGPT in your browser and enable the search function.
But realistically, trying to optimize for every possible AI fanout quickly becomes exhausting. It requires too much manual effort, and don't forget that AI systems like Google want to provide personalized answers.
So you may see something else from your intended audience.
A better approach to ensuring query fan-out coverage is an established SEO practice—topical depth.
Topical authority is about building enough authority around a subject so your brand becomes relevant to a wide range of related questions, irrespective of whether those questions surface in search results or AI-generated answers.
Pages that perform well in AI Overviews frequently rank strongly for either the core query or closely related subtopics.
Surfer’s Topical Map identifies semantically related keywords for your website and groups them into clusters. Each cluster represents an opportunity to reinforce your position within your parent topic.
For example, in the image below, you can see that data governance frameworks have 27 articles for my website. Addressing these missing topics can help strengthen my website’s overall topical coverage.
Since query fan-outs include related searches, I’d suggest starting with a mid-long tail keyword with lower competition.
This way, you’ll increase your chances of ranking in traditional SERPs while still being eligible to rank for the parent topic in Google’s AI Overviews.
You can also filter for lower-difficulty clusters in Surfer’s Topical visual map. Here, I’m looking at potential clusters with low competition.
Once a cluster is selected, you can draft or optimize content within the Content Editor. With query fan-outs, your objective is not to win a single query but rather to create a content network that Google’s AI systems consistently recognize as credible.
2. Write direct, self-contained answers
Google's AI evaluates individual passages on a page to see if they can stand alone as a useful answer. This is known as passage-level extraction, and it gave birth to the concept of an "answer capsule"—a short, precise answer to a query that does not depend on the surrounding content-
An answer capsule is typically placed directly under an H2 phrased as a natural-language question. Here's an example:
Capsule answers make AI engines' job of answering queries accurately much easier, so they're not just favored by AIOs but also other AI platforms. For example, Search Engine Land's research found that 72.4% of pages cited by ChatGPT have answer capsules.
When writing answer capsules, one thing we can carry over from traditional search is to lead with direct answers. But now, it's not about frontloading target keywords but providing complete, self-contained answers at the passage level throughout the article (just like you would for featured snippet extraction).
The best part is that the AI platform you want to get cited in can basically give you the answers you should use in your article. As we're trying to appease Google here, we can ask Gemini a question for which we're writing a capsule and see what comes up, like so:
Google is likely using similar or even the same systems across AI models, so asking Gemini is like Google giving you the answer.
That said, you shouldn't just copy/paste the answer verbatim. It'll probably be pretty long, and you'll see more fluff than with AIOs. It may also violate Google's AI abuse policy.
That's why you should use what Gemini tells you and extract the key points into a short answer. In our example, this could look something like this:
"Vitamin D is important for athletes because it directly regulates physiological processes essential for cardiorespiratory function, bone health, immune resilience, and recovery."
You can then add supplementary information to expand on the topic and demonstrate your expertise.
3. Target long-tail, informational queries
Like other AI platforms and features, AIOs overwhelmingly favor informational, question-based queries. These queries make up around 58% of all AIOs, with long-tail keywords triggering AI Overviews the most.
Specifically, you're looking for queries with 5–7 words on average.
This is the sweet spot because it typically means that the query has more contextual cues than a standard broad keyword, pushing Google to provide a more in-depth response.
You don't need to reinvent the wheel to find the queries you should target. We can do this using many of the same SERP features and tools we used to rely on for traditional keyword research.
For example, you can give a tool like AlsoAsked a broad, generic keyword, and it'll return an entire tree of question-based queries you can target.
You can take an even simpler approach by just searching for the keyword and browsing Google's People Also Ask (PAA) section. As you expand a question, more will pop up to give you more ideas on the conversational queries to focus on.
No matter how you go about it, you should test which of the suggested queries actually trigger AI Overviews.
For example, when I searched for one of the PAA questions, I only got a featured snippet but not an AIO.
But when I used a more definitive one, I got the AIO.
You can do the checking manually or with a dedicated tool, depending on the number of queries you're targeting and your bandwidth. Either way, prioritize queries where you can provide genuinely clear and comprehensive answers that AI engines favor.
As you write answers, use natural, conversational language that mirrors how people actually search and speak. Google search no longer relies on exact keyword matching, so you shouldn't care about it either. Instead of forcing queries in, answer them directly like I explained above to optimize for semantic understanding.
4. Include facts and stats on your pages
We know that Google's AI engine likes facts, but we wanted to figure out how much it cares about them. To do this, we analyzed over 57,000 URLs to see if any patterns would show up.
Unsurprisingly, they did.
We saw that cited pages contained 38% more key facts on average than non-cited pages, and pages with 10+ key facts were cited at more than double the rate of pages with fewer than 5.
AIOs also often recycle the sources they trust the most, and fact coverage directly impacts this trust. The most frequently reused "core" sources in AI Overviews show nearly 2× the fact coverage of pages that were never cited.
The advice here is clear: include as many relevant, reputable stats and facts as possible in your content. It doesn't all need to be first-party research (although it's preferable), but it should be something material that AI engines will find useful and credible.
To ensure your page includes relevant information, use the Facts panel in the right sidebar inside Surfer.
Look for the Optimize for AI Search dropdown, and you’ll find pieces of information that AI platforms deem important. We know this because we analyze the pages showing up for your topic in both traditional and AI-generated answers.
For example, our article on AI visibility tools only has 27 out of 45 important facts. No wonder it’s not showing up in AI Overviews.
Dive deeper into the themes to find exact phrases and sources from which the information is extracted.
The most likely reason why AIOs prioritize facts so much is grounding, which is the practice of connecting LLMs with real-world sources to verify claims. According to Google's documentation, this verification is a key process in Gemini's generation of AI responses, so it's safe to say the AIOs use similar (if not the same) techniques.
You can play to this by enriching your content with data points that encourage citations, such as:
- Survey data from your audience (even small sample sizes beat generic advice)
- Case studies with specific metrics and outcomes
- Industry benchmarks or proprietary datasets
- Expert analysis that adds a new perspective to existing data
5. Structure content for AI extraction
AI engines love well-structured formatting, and AIOs are no different. Our research of over 405,000 searches showed that 78% AIOs contain lists.
You'll probably notice this whenever you search for something. Even if the query doesn't directly indicate a listicle, you'll probably get a list of some kind. Here's an example:
The good news here is that you don't need to learn anything new to adapt your content to AI-powered search—you just need to follow a few well-known SEO best practices.
If you haven't already, write descriptive H2 and H3 headings that clearly signal what each section covers. Avoid vague headings that tell AI nothing, such as:
- "The bottom line"
- "Quick facts"
- "Key details"
Instead, make sure the heading directly summarizes the section under it. And for bonus points, include question-style headings that mirror how users phrase searches whenever you can. For example:
Instead of...
Use:
Ground beef calories
How many calories does ground beef have?
Life insurance types
What are the types of life insurance?
Importance of regular exercise
Why is regular exercise important?
As for the body content, avoid massive chunks of text. Replace or break them up with:
- Short paragraphs (2–4 sentences)
- Lists, bullet points, and tables
- Step-by-step formats where applicable
All of this makes it easier for Google's AI to pick up and cite your content (besides the obvious benefit of making the reading experience more enjoyable for your users).
6. Add schema markup
This is a bit of a roundabout step because schema markup isn't a direct AIO ranking factor. Still, it helps you climb traditional rankings, and our research showed that among the top 3 AIO citations, 54.14% rank in the top 10 when considering the main query or at least one fan-out query.
So by helping you rank in blue-link SERP results, schema indirectly boosts your chances of being cited in AIOs. Besides, structured data helps Google's AI understand what your content represents, so you can only benefit from including it.
This is especially the case with types of schema that can impact general AI visibility, including:
- Article schema (with author, datePublished, publisher fields)
- FAQPage schema (formats content as Q&A pairs, which is exactly how AI extracts information)
- HowTo schema (for step-by-step content)
- Organization schema (for brand and entity identification)
This impact isn't just theoretical—we have quite a bit of proof that schema helps with both traditional search and AIO citations. When Search Engine Land's experiment compared three pages (one with well-implemented schema, one with poor schema, and one with no schema), only the page with well-implemented schema appeared in an AIO and achieved the best organic ranking (#3).
The problem with schema is that implementing it may be a bit complex if you haven't done much technical SEO.
If you need help, Schema.org has a bunch of helpful documentation, and it outlines all the schema types you can use. You can also use the Structured Data Markup Helper if you don't want to implement schema manually.
When you're done, always test your schema using Google's Schema Markup Testing Tool. It checks your code and shows all schema markup on the page, helping you spot and fix any errors.
7. Build brand mentions across the web
A major shift from SEO to any AI optimization strategy (including AIOs) is the focus on mentions over backlinks. Unlike traditional search engines, AIOs don't care as much about who links to you but who mentions you (and in what light).
Ahrefs' study of 75,000 brands confirms this. It found that branded web mentions show the strongest correlation (0.664) with AI Overview visibility, outperforming backlinks (0.218) and domain rating (0.326) by a wide margin. Brands in the top 25% for web mentions get 10x more AI citations than brands in the next tier down.
The takeaway here is clear:
Your website alone most likely won't get you into AIOs, so you need strong off-page signals.
As I mentioned, around 58.1% of citations come from Wikipedia, YouTube, and Google, so sources often receive stronger AI citations than owned domains.
If Google's AI sees a tech review site, a popular YouTuber, and people on Reddit all mentioning your brand as the solution, it triangulates that trust.
There are plenty of ways to build brand mentions, but I won't bore you with all of them—here are some of the fastest and most impactful ones:
- Engage authentically on Reddit: Join the relevant communities and participate in conversations. Make sure not to spam or just talk for the sake of engagement—provide genuinely helpful answers that users are looking for. AI picks up on sentiment, and downvoted self-promotion is a negative signal.
- Create YouTube videos: Film content that answers the same questions as your written content. Google transcribes videos and uses the content as a source, so clear and actionable videos can draw in AIOs.
- Maintain active profiles on industry review platforms: Get on Trustpilot, G2, Capterra, and similar platforms for your industry. Encourage satisfied customers to leave detailed reviews, and respond to any negative feedback honestly and professionally to control the narrative and sentiment.
- Pursue digital PR opportunities: Aim for coverage in authoritative publications. Getting mentioned in industry news, expert roundups, and editorial content builds brand signals that can help Google's AI recognize you as an expert.
8. Continue SEO best practices
It's easy to get caught up in AI-specific optimization and forget the basics. Google Search Central has explicitly stated that there are "no additional requirements to appear in AI Overviews" beyond following standard Google Search Essentials.
This means that the same SEO strategies that drive organic rankings also determine AI Overview eligibility.
Our data from the studies I've mentioned throughout this article confirms this—52% of sources cited in AIOs also appear in the top 10 organic results, and among top-3 AIO citations, 54.14% rank in the top 10.
I'm sure you're already familiar with some of the key SEO practices, but let me just give you a run-through of the most important ones for AIOs:
- Make sure pages are indexable: Don't block Googlebot, and ensure content isn't hidden behind JavaScript rendering that crawlers can't process. You can check for the related issues with Google Search Console's URL Inspection Tool.
- Use descriptive title tags and meta descriptions: Write metadata that clearly communicates the topic. Use direct language that shows both AI and users what the page covers.
- Implement a clean heading hierarchy: Follow a clear architecture (H1 → H2 → H3) so both traditional search engines and AI can follow your content's logical structure.
- Ensure pages load quickly and provide a strong mobile experience: Core Web Vitals are still a key part of Google's page experience signals, so run your pages through PageSpeed Insights to check for speed and usability issues.
How to track AI Overviews
AIO visibility is a new visibility layer, so you should treat it separately from organic traffic and SERP rankings. Monitoring often requires dedicated tools and processes you should include in your overall content strategy.
An easy way to track your AIO visibility is to Google your target query and see if an AIO shows up. If it does, check:
- Whether your site shows up in it
- How it's cited (directly in the answer vs. being included in the sources)
- What position it holds among the cited sources
As simple as this may be, it can get pretty time-consuming. You'd have to routinely run dozens of queries and track everything manually, which becomes a waste of time at some point.
A much more streamlined option is to use a dedicated tracking tool like Surfer AI tracker.
- Enter the name of your brand
- Add your prompts or let Surfer suggest recommended promps to track.
You can then see an overview of your AI Overview's visibility.
But the real power of using a dedicated AI search tool is in the competitor mapping and sources reports.
The Sources report lets you choose competitors and compare your visibility across pages. For example, while Surfer is mentioned on 119 pages along with Ahrefs, 258 sources only mention Ahrefs.
You can also drill down into specific URLs showing up in AI overviews for related prompts. Like this one that mentions 9 brands but not Surfer.
The Competitors report lets me peek into how Ahrefs is doing.
The pages who mention them, like we saw earlier.
And which prompts they show up for. I can use this to edit my prompts if I'm competing for the same queries.
As a side note, avoid tools that use APIs when tracking your visibility in any tool. Our research found only 4% overlap in sources between ChatGPT's API and its web interface, and only 24% overlap in brands mentioned. These are massive discrepancies that can give you a distorted image of your visibility, so always test against the live interface.
Like other AI features and tools, AIOs constantly evolve. The algorithms and citation patterns will shift, so starting optimization and monitoring early helps you evolve ahead of the curve and build a compounding competitive advantage.
Turn these tactics into AIO visibility
AIOs, AI Mode, and other new features aren't replacing SEO—they're just building on it. If your pages don't follow the established SEO strategies, no amount of AI-specific optimization can make up for that.
AI Overviews appear for plenty of queries, so there's a lot you can do to optimize for the most relevant ones. As you've seen from both our and third-party research, all the data points to a few universal predictors of AI citations.
Here's a quick refresher alongside the concrete steps you should take:
.avif)
