How Google’s AI Overviews Choose Sources (and How to Become One)

Search for just about any informational keyword on Google now, and you’ll probably see an AI Overview.

While AI Overviews have been a thing since May 2024, their prominence has gone through the roof.

According to a recent study by Xponent21, more than 50% of all searches on Google generate AI Overviews.

Also, since they appear at the very top of the page, they’re occupying prime SERP real estate that was once reserved for Featured Snippets and the top-ranked organic results (the ‘blue links’).

The result?

Brands appearing as sources in AI Overviews are capturing the majority of users’ eyeballs.

Not only do they benefit from the increased exposure, but evidence shows that getting listed as a source in an AI Overview can improve your CTR by 80% (to 1.08% from 0.6%).

If you aren’t listed as a source in an AI Overview, your CTR will likely take a nosedive.

Research from Ahrefs found that the presence of an AI Overview can reduce clicks by 34.5%, so you have every incentive to get cited in AI Overviews.

This guide will teach you:

1. Which signals influence AI source selection

2. Common patterns in frequently-cited content

3. How to optimize your content for AI Overviews

What are Google’s AI Overviews? How Do They Impact Organic Visibility?

Google first began using AI with RankBrain in 2015, but it’s only a machine learning component that helps its algorithm interpret queries and deliver relevant search results (which it still does, alongside more advanced models like MUM and BERT).

However, the announcement of Google’s Search Generative Experience in May 2023 marked the first time the company hinted at using generative AI to enhance its search results. It also sent the SEO world ablaze with anticipation (and fear).

At first, SGE was an experimental Search Labs feature. It remained this way for a full 12 months until it was rebranded as AI Overviews (AIOs) and went live for all US users in May 2024.

Here’s what one looks like for reference:

AIOs were highly prominent right off the bat, and they appeared for approximately 40 - 50% of all search queries.

This move was met with considerable backlash, though, primarily due to inaccuracies (like the infamous glue-on-pizza incident).

In June 2024, Google would roll AI Overviews back so that they didn’t appear as often. Instead of generating for half of all user queries, they shot down to 15% in June.

Google would reduce the prominence of AIOs even further in July, when they hit their lowest point at 8 - 9%.

As we’ve already covered, they’ve since reclaimed their throne, as AI Overviews now appear for 50% of search queries again. This time, though, they’re powered by an upgraded version of Gemini.

It didn’t take long for SEOs to notice that AI Overviews were reducing CTRs whenever they were present.

AIOs have also caused zero-click searches to skyrocket.

According to research, 60% of searches in 2024 ended without a click.

This is a concerning statistic, but as evidence is starting to show, zero-click searches can still:

• Introduce your brand to target audiences

• Improve brand recall for branded searches and direct visits later on

• Build trust and spark interest

According to one brand, they saw a 30% decline in their CTR on Google, but their search impressions rose by 49% in that same period (Ridge Marketing).

They’re certainly not alone, leading some marketers to notice that ‘clicks are down, but leads are up.’

What are the components of an AI Overview?

Next, let’s examine each aspect of an AI Overview to learn more about how they work.

For the purposes of this example, we’ll look at the AIO that generates for the keyword ‘what are LLMs’ on Google:

It’s important to note that AIOs are now abridged at first, meaning that users must click on the Show More button to see the full Overview. Because of this, #1-ranked organic results, ads, and other SERP features are sometimes visible without needing to scroll down.

This is in stark contrast to when AIOs first rolled out, as they dominated the entire results page, forcing users to scroll down to find organic results and ads.

Due to criticism, Google quietly compressed its AIOs to make them less intrusive.

If users do expand the AIO, here’s what they’ll see:

At the very top of the page, Gemini generates an answer to the user’s query.

This takes the form of a brief, bolded snippet that provides answers, definitions, and other quick factoids. Most of the time, users will find the information that they need here and end their search experience (resulting in a zero-click search).

However, notice the prominent source panel on the right-hand side.

It contains the three primary online sources that Gemini used to generate its answer. These source panels are now prime SERP real estate because they:

• Improve click-through rates

• Introduce users to your brand name (even if they don’t click)

• Boost things like brand recall, branded searches, and direct site visits

The AIO also provides a helpful YouTube video, which it often does for informational queries. That means producing videos for informative keywords is also a good way to improve your visibility in AIOs.

Beneath the YouTube video, the AI Overview actually keeps going:

Here, Gemini provides a more detailed breakdown for the user. In this case, it gives us more information about how LLMs work.

Notice the tiny hyperlink symbol next to each nugget of information.

These are more online sources, and clicking on them will reveal organic results to users. While getting included as a source here is nice, getting listed in the source panel is the real goal.

At the end of the AIO, users have the option to transition into AI Mode. This button prevents another challenge for marketers since it can take users away from the SERP (and stop them from exploring the organic results).

It’s another reason why getting listed as one of the top 3 sources in an AIO is such a big deal now. The source panel is the cream of the crop when it comes to SERP visibility in the current era.

What Signals Influence AI Source Selection?

Next, let’s analyze how Gemini decides which sources to include in its AIOs.

Before we get started, it’s crucial to mention that Google’s exact source selection process is proprietary and not shared with the public. This is nothing new to search marketers since Google’s official algorithm is also kept under lock and key.

However, that hasn’t stopped us from cracking their search algorithm’s code, and their LLM is no different.

Bearing that in mind, these are the trust signals we were able to infer from:

• How large language models work in general

• What Google has stated publicly

• Google’s existing trust signal infrastructure (like E-E-A-T)

Let’s take a closer look.

Existing trust signals and content quality

Gemini is a Google product, so it makes sense that it would lean on the company’s existing trust ecosystem.

This means Gemini pays attention to:

• E-E-A-T (experience, expertise, authoritativeness, and trustworthiness) signals. The E-E-A-T quality rating system also applies to Gemini, so you need to create helpful content that showcases first-hand experiences and expertise.

• Topical authority. Sites that have topical authority are preferred by LLMs, meaning they’ve established notable expertise on a particular topic (or topics). A great way to achieve this is to publish content clusters (interlinked pieces based on keywords covering the same topic).

• Authoritative, contextually relevant backlinks. Yes, backlinks still matter in the age of AI search, especially to Google. Yet, LLMs do not measure ‘domain authority’ by using third-party metrics (like DA and DR). Instead, contextually relevant backlinks (especially ones that mention your brand) are what matter most.

Reputation also matters because LLMs will check author bios for signs of experience and expertise (credentials, awards, keynote speaking events, etc).

Semantic relevance to the query

LLMs like Gemini are able to read and actually comprehend user prompts, which means they don’t have to rely on exact-match and partial-match keywords.

Instead, semantic relevance to the query is far more important than a page that has a matching key phrase.

For instance, an LLM like Gemini may cite a piece of content that doesn’t contain any target keywords from the prompt. If the source is semantically relevant (i.e., it provides helpful information for the user query) and comes from a reliable domain (like a .gov or a trusted niche blog), it’s fair game for an AI citation.

Concise content formatting and structured data

Because LLMs have to summarize multiple sources at once, they prefer content that has concise formatting.

Headings, bulleted lists, FAQ sections, and HowTos are some of the things LLMs like to see when compiling sources.

Pairing this with semantic HTML and schema markup will make your formatting even more noticeable to Gemini (and other AI search tools). That’s because structured data like schemas make your content machine readable by clearly labelling things like FAQ pages, recipes, and reviews. You can find a full list of accepted schemas at Schema.org.

Content freshness

AI companies always receive scrutiny whenever their LLMs provide false or outdated information. That’s why lots of AI search tools have a ‘recency bias’ of sorts.

In other words, these AI tools want to make sure they’re providing users with the most accurate, up-to-date information on any given topic. This especially applies to time-sensitive topics like news stories and consumer trends (which change all the time).

The best thing you can do is periodically update all your non-evergreen content so that it’s as fresh as possible.

Brand citations

LLMs like Gemini place a lot of importance on brand citations, even if they don’t contain backlinks. Once again, this is because they’re able to actually read and comprehend brand mentions, so a link isn’t necessary for it to improve your visibility (although it helps traditional search, which powers AI search).

The quality of citation matters more than the quantity, though. LLMs want to see your brand appearing in authoritative, niche relevant content, not random websites.

This is where having a high-quality link-building strategy in place really pays off.

If you’ve already been building backlinks on authoritative websites, your brand is likely already leveraged to appear in lots of AI searches.

How to Optimize Your Content for AI Overviews

Now, it’s time to learn how you can optimize your content to improve your visibility in AIOs.

These are modern tactics that consider what LLMs value in content, so they don’t rely on irrelevant SEO techniques of old.

Content clustering to build topical authority

If you want Gemini to recognize that your site has topical authority, you need to start creating content clusters.

A content cluster consists of:

• A seed page. This page introduces the general topic and links out to all the cluster pages. For instance, a content cluster about digital marketing could have a seed page called ‘The Ultimate Digital Marketing Guide for 2025.’ From there, you’d build a series of cluster pages that dive deeper into related subtopics.

• Cluster pages. These pages explore topics related to the seed page. For the digital marketing example, you could have cluster pages that cover things like social media (‘Best Times to Post on Instagram and X’), content marketing (‘How to Create Blog Posts That Attract Attention’), and analytics (‘How to Measure AI SEO Success’).

The idea is to cover every angle of a topic by creating a series of related pieces that all link back to the seed page, and vice versa. Internal links are crucial for successful content clusters, so make sure your clusters are interlinked.

LLMs and search algorithms will notice that you consistently provide detailed, accurate information for topics related to your products and services, and topical authority will begin to form.

Digital PR to generate more brand mentions

From Gemini to ChatGPT, AI-powered search platforms use brand mentions as trust signals. The best way to earn more authoritative brand mentions is to engage in digital PR techniques.

Machine-readable formatting and high content standards

As we mentioned, including structured data in your content makes it machine-readable. Semantic HTML and schema markup are the two most important types of structured data for AI search.

The most important schema types include:

• author

• FAQPage

• review

• aggregateRating

• article

• sameAs

Monitor your visibility using up-to-date tools

Lastly, you can’t improve your visibility in AIOs if you have no way to measure your success. The good news is that many marketing analytics platforms are finally including AI tracking capabilities.

To track your visibility in AIOs, you can use:

1. Ahrefs

2. Semrush

3. Keyword.com

4. BrightEdge AI Catalyst

Before we wrap up, here’s a table distinguishing AI-friendly SEO techniques from tactics that don’t improve AI visibility:

Closing Thoughts: Google’s AI Overview Source Selection

In short, the introduction of Google’s LLM-powered AI Overviews has revolutionized the search marketing game.

Appearing as a source in AIOs is now critical for improving online visibility.

Focusing on things like top-tier brand mentions, content quality, answer completeness, and structured data will get LLMs to start noticing your brand.

Do you need help measuring the effectiveness of your link-building strategy, especially considering the impact of AIOs?

Book a call with our team of experts to learn where you stand (and how you can achieve your business goals through link-building)!