Understanding how to optimize for AI Overviews can help you capture significantly more space on the search results page. This is becoming increasingly important as AI Overviews occupy a large amount of real estate, often pushing organic results much lower on the page—especially when users click Show more.

Since their release in May, many have been wondering: How exactly can you optimize for AI Overview visibility? Does it differ in any way from standard SEO practices?

We do have some clues to start with. The patent for Google’s AI Overview system offers insights into how search result documents are selected and linked. In my initial analysis, I examined how the system works and answered the question, How does Google AI Overview work? The original version of this article focused solely on how documents were selected based on the patent. Since then, I published several articles covering a variety of aspects of the AI Overview system, including the role of Google Gemini, related queries, as well as collaborative studies with Authoritas that assess the role of links, related queries, user intent, and industry-specific trends. With these expanded insights, I am revisiting and refining the strategies discussed here to help you optimize your content to secure links in AI Overviews.

In this updated article, originally published on June 11, I will provide an overview of how Google’s AI Overview system evaluates and selects content, outline the measures that impact visibility, and discuss strategies for optimizing content based on recent findings and the insights from Google’s patent. My goal is to equip you with practical guidance to enhance your content’s visibility and position within AI Overviews.

AI Overview Document Selection and Linking Review

The AI Overviews system seeks out search result documents (SRDs) at two different points in its process. First, during the initial creation of the summary, then again at the end to add links.

Initial Search for Documents Responsive to Direct Match Queries

The system begins by evaluating documents directly related to the user query, focusing on factors such as ranked position, uniqueness, quality, and trustworthiness. This initial evaluation prioritizes documents that meet these high standards to ensure that the most authoritative and relevant content is included in the summary. To improve your content’s chances of being chosen, focus on enhancing its quality, credibility, and alignment with user intent.

Leveraging Related and Reformulated Queries

If the initial documents do not meet the required standards, the system broadens its search to include related and reformulated queries. By doing so, it diversifies and verifies its sources to create a more comprehensive summary. Including content that addresses related topics and keywords helps expand the range of documents that may be selected. Our findings showed that addressing these additional queries significantly increased the percentage of links included in AI Overviews.

Verification and Embedding Distance

When adding links, the system re-evaluates documents to verify the accuracy of the summary. It looks at the embedding distance between the content in the summary and the related content in the document, assessing how semantically similar they are. This is a critical factor that influences whether a document will be linked as a supporting source.

In the example below, you can see that the statement “Ducks have symmetrical bodies, metallic feathers, and move with a liquid grace” is verified by the highlighted text in the linked article, “With their vivid, metallically colored feathers, their perfect symmetry and liquid movement, ducks are beautiful by every metric (with the obvious exception of their voices).” This is accomplished through embed distance.

Role of Google Gemini in Selection

The Large Language Model (LLM) underlying the AI Overview system, Google Gemini, plays an important role in document selection. This model uses retrieval-augmented generation (RAG) to pull relevant information from Google’s index, training data, and the knowledge graph. By aligning your content with how LLMs retrieve and analyze sources, you can enhance its chances of being selected. Ensure your content is comprehensive, authoritative, and structured to match user intent.

Content Relevance and Multimodal Integration

Google Gemini’s multimodal capabilities allow it to process text, images, and other formats, creating fuller, more context-aware summaries. Incorporating multimedia elements, such as relevant images or YouTube videos, can make your content more appealing during both the summary creation and verification phases.

What follows are the specific measures highlighted in the patent that are potentially used during summary creation and linking.

For a more in-depth understanding of how the AI Overview system works, read my article How Does Google AI Overview Work? Insights From the Patent.

Document Evaluation Measures in Google’s AI Overview Patent

The Google patent for the AI Overview describes several measures used to evaluate documents for inclusion in the summaries. The measures fall into the following categories:
Query-Dependent, Query-Independent, User-dependent, Related Queries and Contextual Relevance, and Context and Profile Data Utilization.

Query-Dependent Measures

When a user enters their query into the Google search bar, the system evaluates search result documents (SRDs) using several query-dependent measures to assess their relevance and suitability.

POSITIONAL RANKING

Positional ranking is one of the primary factors the system considers when evaluating SRDs. The system also limits how deep into the results the system will go to select documents. While this measure is called out only a couple times in the patent, this mention highlights the importance of ranking highly for gaining visibility in AI Overviews.

Selection Rate

The selection rate, essentially the click-through rate (CTR), is another measure referenced in the patent. Despite Google’s historical stance denying CTR as a direct ranking factor, the recent leak and the mention of this metric in the AI Overview patent imply that user engagement metrics do influence how content is evaluated. This highlights the need to optimize for user interaction with compelling titles, headers, and meta descriptions.

Locality Measure

The locality measure looks at where the query was made and matches it with the location relevant to the search result document. This helps ensure that it’s responses are geographically relevant to your location. For example, if you search for “best pizza places near me,” the system can prioritize results that are near your current location. This makes the search results more practical and useful, helping users find local businesses and services that meet their immediate needs. As of this writing, localized “near me” results for these types of queries are not in play.

Language Measure

The language measure compares the language of the query with that of the SRDs to ensure that the content matches the user’s language preferences, improving the relevance and usability of the results.

Query-Independent Measures

The system also evaluates documents using query-independent measures that assess overall document quality and promote diversity in the selection process.

Selection Rate for Multiple Queries

A high selection rate across multiple queries signals that a document is broadly relevant and trusted by users. This measure, basically a cross-query CTR, indicates consistent user interest and relevance, improving the document’s chances of selection.

Trustworthiness of the Document

Straight out of Google’s EEAT guidelines comes Trustworthiness. To evaluate the trustworthiness of the document, the patent specifically mentions 3 criteria:

• Author: The credibility and expertise of the content’s author.
• Domain: The overall reputation and authority of the website hosting the content.
• Inbound Links: The quality and quantity of links pointing to the document from other reputable sources.

Although specific evaluation metrics like domain age are not mentioned, the focus on trustworthiness (mentioned 9 times in the patent) spotlights the importance of author credibility and authoritative backlinks.

Overall Popularity

While the patent does not detail the exact methods for evaluating overall popularity, it suggests that content popular in organic search and broadly engaged with across the web is favored. Metrics like social shares, user interactions, and traffic can all contribute to this measure.

Freshness of the Document

The system prefers content that is current and frequently updated. Freshness is assessed based on the date of creation and the recency of updates, which aligns with the need to keep content relevant and timely.

Diversity Among the Selected Documents

The AI Overview system seeks diversity among selected documents to present a range of perspectives and content types. This approach prevents redundancy and ensures that users are provided with varied and comprehensive information. This can benefit sites that may not dominate top rankings but offer unique insights or alternate formats. This measure is very important as it pertains to the documents linked to in AI Overviews that don’t overlap with the organic search results for the query. This ties into why the system seeks out documents responsive to related queries as noted earlier.

User-Dependent Measures

To provide personalized and contextually aware responses, the system leverages several user-dependent measures.

User Profile Attributes

The system may prioritize content that matches user interests and preferences, based on their profile. For example, a fitness enthusiast might see results related to workout routines and nutrition tips, ensuring the content is relevant to their specific interests.

Overlap with Recent Queries

This measure assesses how closely related the current query is to the user’s recent searches. If there is significant overlap, the system will prioritize documents that continue the line of inquiry, such as a user who searches for “best smartphones 2024” and then “smartphone reviews.”

Recency of Queries

The system also considers how recent similar searches were made, prioritizing content that matches ongoing interests. For instance, a search for “Italian restaurants” followed by “best pasta dishes” the next day suggests continued interest in related topics.

Interactions with SRDs

The system evaluates how users interact with previous search result documents. If a user frequently engages with specific types of content or sources, the system will prioritize similar documents in future queries.

Profile Data

Beyond direct queries, the system may analyze a user’s broader online behavior, such as browsing history and social media interactions, to enhance content selection. This allows for a more tailored search experience, such as prioritizing travel content for a user who frequently reads travel articles.

These measures ensure that the selected documents are tailored to the user’s interests, recent activities, and behavioral patterns, thus improving the relevance and quality of the search results.

Related Queries and Contextual Relevance

When top-ranked documents do not fully meet the criteria, the system searches for documents responsive to related queries. This process helps expand the pool of relevant content and improve overall coverage.

Correlation Threshold

The magnitude of the correlation between the input query and related queries must meet a specific threshold, often measured through embed distance. This threshold indicates the strength of the relationship between queries. In AI Overviews, the system prioritizes queries that are highly similar or contextually relevant to the initial query. If the relationship exceeds the threshold, related documents are deemed more relevant and are more likely to be included in the overview.

Topical and Entity Overlap

Assesses the degree of topical and/or entity overlap between the current query and recent queries. This measure ensures that the system identifies and prioritizes documents that cover similar topics or entities (like people, places, or things) as those recently searched by the user. This helps maintain consistency and relevance in the search results. For example, imagine a user searches for “climate change effects in coastal cities.” Shortly after, they search for “rising sea levels impact.” Due to topical and entity overlap, the search system recognizes that both queries are related to climate change and coastal regions. This measure ensures that the AI Overview prioritizes documents that consistently cover similar topics, such as environmental impacts and geographic locations. As a result, the user sees search results that are relevant and consistent with their recent searches, offering a more cohesive understanding of the broader topic.

Temporal Proximity

The quantity of occurrences of the query and the related query both being issued by a corresponding device or account within temporal proximity of one another. This means the system checks how frequently the initial and related queries are made within a short time frame. High temporal proximity suggests that the user is interested in related topics during a specific period.

Amount of Time Passage

Measures the temporal proximity between the current query and recent queries. This measure helps the system understand how recent the user’s previous queries were in relation to the current query. If the recent queries are close in time, it indicates that the user’s interest in a topic is still relevant and ongoing.

Context and Profile Data Utilization

Automatic Generation of Implied Queries

The system can generate implied queries based on context and profile data, enhancing the search results’ relevance. For example, a query for “best hiking trails” could prompt related content on “hiking gear” or “weather in hiking areas.”

Context and Profile Overlap

The system uses profile data and recent activity to determine if an implied query aligns with current interests. If a strong overlap is detected, the system prioritizes documents that best match the user’s profile and search context. If a user frequently searches for “beginner hiking trails” and “hiking gear for beginners,” and then searches for “best outdoor activities,” the system recognizes an overlap with their interest in hiking. Based on this context, the AI Overview prioritizes results focused on hiking as an outdoor activity, aligning with the user’s profile and recent search behavior.

AI Overview Measures Reflect Google’s Ranking Factors

The visibility measures described in the patent above, are likely providing insight into how documents are evaluated within Google’s index and ranking systems, creating the foundation for what the AI Overview system ultimately selects and displays. Rather than searching the internet in real time, the AI Overview accesses the pre-evaluated and top-ranked documents from Google’s Index and ranking systems.

Let me say this more another way…it makes no sense for AI Overviews to re-evaluate documents. It would take up more computational resources which the system seeks to minimize. Google is likely relying on it’s other systems to evaluate quality.

Key visibility factors—such as search position, selection rate (CTR), authorship, trustworthiness, content popularity, and freshness—mirror standard SEO priorities. High-ranking documents have an advantage, with additional emphasis on diverse sources and engaging titles to attract clicks. Trustworthiness is further supported by author credibility, domain reputation, and quality inbound links, aligning with Google’s EEAT (Experience, Expertise, Authoritativeness, Trustworthiness) standards.

Overall, these measures highlight the importance of ranking, engagement, credibility, and relevance, reinforcing that SEO practices support visibility in both organic search and AI Overviews.

Now that we have a good handle on the theoretical side of how AI Overviews work based on the patent, let’s see what research has shown us so far.

What We’ve Learned About Optimization Based on Subsequent Research

Through research into AI Overview link selection, several important insights have emerged, informing our understanding of optimization for visibility. My joint efforts with Authoritas have not only supported our initial hypotheses but also revealed new perspectives on how Google’s AI Overview system selects and links to content. Here, I’ll highlight some of the lessons learned.

Emphasis on Informational Content

A significant takeaway from Authoritas’s and my AI Overview intent study is that informational queries most frequently trigger summaries, aligning with their current focus on providing knowledge-rich responses. This was evident in our research, which found that 28.8% of Informational queries generated AI Overviews, followed by Commercial: Informational queries at 18.6%. This focus highlights the importance of creating comprehensive, well-researched content that meets informational needs to enhance visibility in AI Overviews.

Top SERP Positions and Inclusion Probability

Our research found a 46% overlap between links in AI Overviews and top ten organic rankings. Further, the top two SERP positions had the highest likelihood of being included in AI Overviews. Content ranking in the top position had a 53% chance of appearing in an AI Overview, while the second position followed closely behind. This trend indicates that securing a top-two spot significantly boosts the chances of content being featured, not just in organic results but also in AI-generated summaries.

If 46% of AIO links overlap with the top organic results, what about the other 54%?

This is where related queries enter the equation!

Inclusion of Related and Reformulated Queries

Our findings further provided support for what we learned from the patent: AI Overviews not only link to documents responsive to direct match queries but also incorporate related and reformulated queries to diversify and verify their content. When related and reformulated queries were included, the percentage of links accounted for in AI Overviews rose from 46.3% to over 67%. This pattern shows that Google’s AI Overview system draws from a broader range of query results to ensure comprehensive and reliable summaries.

Domain Authority and Frequent Sources

The analysis of frequently cited domains by intent revealed a clear pattern: authoritative and niche-focused sites often hold the advantage. YouTube, Wikipedia, and Investopedia were leading sources for informational queries, indicating that content from well-established sites is highly valued in AI Overviews. For brands and content creators, this points to the necessity of building domain authority, fostering credible inbound links, and maintaining expertise in their content areas to improve visibility.

Industry Variations in Document Selection

We found notable industry-specific patterns in AI Overviews document selection. Industries such as Health and Education, which often demand detailed and reliable information, see a higher average number of links in their AI Overviews, reflecting the need for comprehensive sources. In contrast, industries with less complex informational needs, like Entertainment, include fewer links. This implies that content creators in highly informative industries should aim to provide resource-rich, authoritative content to maximize their visibility in AI Overviews.

Key Pathways to AI Overview Visibility

Taken all together, what we know from the patent and our research so far, this creates several pathways to visibility in AI Overviews. I highlight below 3 unique pathways and 3 sub-pathways to gaining visibility in AI Overviews. The pathways are as follows:

Each pathway is tailored to different ranking scenarios, content quality, and query competitiveness. It all starts with SEO.

SEO as the Foundation for AI Overview Visibility

To gain visibility in AI Overviews, strong SEO is required. Ensuring your content is optimized for Google provides the foundation needed for inclusion in these summaries. This includes optimizing technical SEO, conducting comprehensive keyword research, and aligning content with user intent. With these SEO fundamentals in place, you can pursue the following specific pathways to achieve AI Overview visibility.

Pathway 1: Direct Match Queries

To improve your content’s chances of being selected in AI Overviews, aim to rank in the top 10 for your primary targeted query. Focusing on direct match queries means aligning content closely with specific, high-intent searches. Comprehensive keyword research and user-intent alignment can help you secure high rankings, which is essential for achieving direct AI Overview inclusion.

Pathway 2: Target Related Queries

In addition to targeting direct match queries, you can fortify your relevance by targeting related queries. Even more important, if the top 10 rankings for your primary query are highly competitive, you can target related but less competitive queries to gain visibility. Covering associated topics through a topic clustering strategy signals to AI Overview algorithms that your content is contextually relevant. This approach can improve your chances of inclusion even if you’re not ranking for the main query, as related queries offer an alternative route to AI Overview selection.

Shared Sub-Pathways for Direct and Related Queries

• Top 2 Ranking: For both direct match and related queries, securing a top 2 ranking significantly increases the likelihood of appearing in an AI Overview. These high positions indicate to the system that your content is highly relevant and authoritative on the query topic, maximizing visibility potential.
• Unique, High-Quality Information: If you’re already ranking in the top 10 but still not appearing in AI Overviews, enhancing your content with unique insights, original research, or specialized perspectives can set it apart. Offering exclusive value positions your content as a more desirable option for AI Overview algorithms.
• Match Embedding Distance: The AI Overview system evaluates how closely your content’s language and structure align with the summary information it generates. To improve semantic relevance, use consistent terminology and structure that match the tone and details of AI Overview content, thereby increasing your chances of selection.

Pathway 3: YouTube

YouTube provides an alternative route to AI Overview visibility. Creating optimized, relevant video content can help attract attention from AI Overviews, which currently value YouTube content. Having an optimized YouTube channel with relevant videos can enhance user engagement, appeal to a broader audience, and boost the likelihood of selection.

Bringing It All Together for AI Overview Visibility

Each pathway offers a unique approach to gaining visibility in AI Overviews—whether through direct rankings for targeted queries, covering related topics, providing unique content, aligning with embedding distance, or leveraging multimedia. By assessing where your content currently stands and selecting the appropriate pathway, you can enhance its potential for selection in AI Overviews, ultimately boosting your search visibility and engagement.

In addition to pursuing these pathways, prioritize Informational Intent in your content. AI Overviews are most often triggered by queries with informational intent, so focusing on comprehensive, well-researched answers that address user questions directly can significantly increase your content’s relevance and visibility.

Finally, stay informed about Industry Trends that may affect AI Overview inclusion. Certain industries—such as Health, Finance, and Education—are more likely to appear in AI Overviews, often due to high informational demand and search volume. By tailoring your content to meet the needs and trends within your industry, you align it with the types of queries that are more frequently selected for AI Overviews.

By combining SEO foundations, targeted pathways, and a focus on informational intent and industry relevance, you can position your content effectively for selection in AI Overviews, ultimately enhancing its reach and impact.

Lastly, keep in mind that the AI Overview system is not static. The algorithms will change and the underlying LLM, Gemini, will change what it selects for its sources to create the summary and what it link to.

Final Thoughts

Optimizing for AI Overview visibility involves a combination of established SEO best practices and targeted strategies informed by recent research and Google’s patent insights. While high SERP positions and direct query matches are quite important, expanding content to include related can provide significant advantages. Emphasizing user engagement, EEAT principles, and content freshness aligns your strategy with both traditional SEO and AI Overview requirements.

Ultimately, achieving visibility in AI Overviews means producing content that is authoritative, comprehensive, and contextually relevant. The system’s evaluation, which considers factors like embedding distance and topical depth, highlights the need for an approach that goes beyond basic optimization. By integrating these strategies and continuously adapting to industry trends, content creators can strengthen their presence in this evolving search feature, positioning their content for maximum reach and impact.

For additional insights on optimizing for Google’s AI Overview through reverse engineering, check out my article on how to optimize for SGE.