ai overviews

The Ultimate Guide to Google AI Overviews

Last updated September 22nd, 2025

Table of Contents

What are AI Overviews?

AI has been integral to Google’s ecosystem for over a decade. More recently, with the proliferation of generative AI, Google is integrating this innovative technology into multiple products across its product line. In search, in the second half of 2023 Google embarked on a year-long experiment in labs called Search Generative Experiences (SGE). During that time Google tested and learned how generative AI could be most effective for search. We monitored that activity with BrightEdge Generative Parser™ (BGP) which we published in our Ultimate Guide to Google SGE.

On May 14th 2024, Google announced that Search Generative Experiences (SGE) would become AI Overviews (AIO) and start appearing for users outside of labs. AI Overviews are built on the same Google Gemini AI Language Model as SGE, and BGP can track their presence just as it did with SGE. AI Overviews are available to US users who are logged into Google, regardless of whether they’ve enrolled in Google Labs. However, as of this writing, they are still not available to non-logged-in users. This development represents a significant step forward for AI-powered search and has major implications for search engine optimization (SEO). Now, SEO must account for traditional organic search results as well as the citations that AI Overviews use to answer questions.

At BrightEdge, our mission is to help marketers understand how these innovations from search engines will impact how they connect with their customers. This guide serves as your resource for AI Overviews, equipping you with the knowledge to understand how generative AI technology affects how your audience interacts with digital content.

How do AI Overviews appear?

Presently, AI Overviews are limited to the US and appear for users who are logged into Google. However, Google has confirmed it is showing AI Overviews to users who are not logged in during small tests.

AI Overviews consist of short paragraphs, bulleted lists, images, and product carousels that Google generates using Generative AI to add context and summarize search results. They appear at the top of the search results, below most paid and above organic listings. To date, AI Overviews appear in a collapsed state, requiring the user to click to read the full content:

When the user clicks "Show More", they are taken to an experience where different modules provide greater detail on the answer. Citations for where the AI Overview obtained the information to compose its content are also shown throughout the experience. In this case, we see that TYR, RunRepeat and the American Academy of Podiatric Sports are cited to for the Generative AI’s content:

The degree to which AI Overviews appear for a user can be controlled (although not entirely turned off) in the Labs section of Google. By enabling AI Overviews, it will deliver fuller, more comprehensive answers. Google has confirmed that turning this feature off will not eliminate AI Overviews from the experience entirely.

For users in the US who are logged in, AI Overviews are currently appearing in less than 15% of all queries. For marketers who need to understand how AI Overviews are appearing for their keywords, it is important to ensure you’ve enabled these results in Google Labs to see the most comprehensive version that your audience may see. Based on Google’s deployment, it is also crucial to identify which keywords are most likely to trigger an AIO.

The presence of AI Overviews varies based on the industry. Data suggests this correlation is largely due to the nature of the industry. For example, in healthcare or B2B technology, where queries are more informational in nature, there is a significant difference in presence compared to industries like restaurants or travel, where user behavior is more transactional (booking reservations, finding addresses, calling the restaurant). This observation holds in other industries as well, except for finance, where Google is not deploying AI Overviews frequently.

When do AI Overviews (AIO) Appear?

AI Overviews are largely appearing for informational queries. Fortunately for marketers, there are signals from traditional search features that can be used as a proxy to estimate where an AIO is likely to appear. According to BrightEdge Generative Parser™, there are correlations between SERP features and the presence of AIO.

As displayed here, featured snippets have the strongest correlation with the presence of an AI Overview. This makes sense because featured snippets mostly appear for informational queries. Conversely, certain navigational and transactional elements have a negative correlation with the presence of AI Overviews. Local 3 Packs, which are deployed for local transactional searches, have a negative correlation with the presence of AI Overviews.

In addition, we see that the presence of sitelinks, which typically indicate that the intent in the query is navigational or a branded term, which is an area where Google is working to reduce the potential for a hallucination.

For marketers who need to detect where AI Overviews could be appearing for their keywords, it will be essential to track the elements in a traditional result that correlate with the presence of an AI Overview. To do this, advanced capabilities such as Data Cube X in BrightEdge allow users to filter results by specific search features and leverage exact match volumes, so marketers can differentiate how long-tail queries, misspellings, and plural/singular versions of keywords may elicit different results.

To begin sensing where AI Overviews will appear, BrightEdge customers should use Data Cube X and filter keywords that contain the following features:

SEO Action Items for When AI Overviews Appear:

  1. Identify Query Intent: Keywords that are informational are more likely to feature an AI Overview than transactional or navigational queries. Use filtering to identify the questions and top-of-funnel keywords related to your brand.
  2. Leverage Traditional Search Features to Pinpoint Where AIO Occurs: BrightEdge users can create keyword groups to track where featured snippets are appearing in search results. These should be monitored and targeted for optimization.
  3. Create Content that Addresses the Questions AI Anticipates: Given that AI Overviews are appearing for informational queries and anticipate follow-up questions, marketers need to build content with this framework when targeting keywords where an AI Overview is likely to appear. To do this, content briefs need to address questions related to the keyword or topic being targeted. BrightEdge users can leverage Copilot for Content Advisor to do this faster. It identifies related questions to the topic, scans content currently ranking, and creates briefs and first drafts that address the topics needed to appear in both traditional ranking and AI Overview citations.

How are AI Overviews formatted?

Within the various states of AI Overviews, a range of content modules are deployed and tailored to match the user's search intent. These content modules, while diverse in their specific applications, exhibit shared structural characteristics, adapting to the unique nature of each search query.

In describing and categorizing these content modules, we have adhered to the terminologies and descriptions provided by Google, ensuring accuracy and consistency in our analysis. This approach helps delineate how AIO tailors its responses to different search intents, providing valuable insights into the operational mechanics of AIO and its impact on search results.

Unordered Lists

Unordered Lists are utilized in two distinct styles, each serving a specific purpose in presenting information:

Style 1- Standard Bullet Point List: This style is straightforward, featuring a series of bullet points. It is used to present information in a simple, easy-to-scan format, ideal for quick comprehension by users. This format is particularly effective for listing features, benefits, or various options related to a search query.

Style 2- Enhanced List with Breakouts: This style includes additional elements such as expanded descriptions or interactive features within the list. It's designed to provide more detailed information and encourage user interaction. This format is suitable for queries where users might benefit from more comprehensive details or where engagement with the content is crucial.

In addition to unordered lists, ordered lists play a small role in AI Overviews appearing across a wide array of informational queries and spanning various industry types. The manner in which these lists are rendered is closely tied to the specific content they are associated with, reflecting the diversity and adaptability of AIO in presenting structured information.

Key SEO Considerations for Optimizing Unordered and Ordered Lists:

Product Viewers:

To aid shoppers in understanding different products, AI Overviews offers a couple variations of product viewers which orient the products to what the AI recommends. Based on the type of search, there are a few ways products can be presented in AI Overviews:

Standard Product Viewer

A standard product viewer offers multiple images of products with pricing and other pertinent details that align to the query intent.

Product Carousels

Product carousels offer less images per product but more shopping detail such as pricing and reviews. These occupy more real estate on a search result page as they offer more products but help the user shop through a wider variety of products.

SEO Implications for Product Viewer Modules:

Baseline AI Overview Components

AI Overviews, regardless of the specific state or module in use, incorporates a set of standard features in its results. These features are consistent across various queries and provide a structured approach to presenting information.

Answer Box:

Source Panel:

Special Cases

As Google progresses in refining AI Overviews, specialized modules are appearing for small sets of queries. Specific elements, tailored to distinct query types, are being introduced. These elements are designed to either integrate cohesively with or complement the existing search modules.

Pro/Con

For a small number of queries, we detect pro/con boxes that are unordered lists that weigh pros and cons of products.

Combined Shopping Offer Cards

For certain e-commerce queries, BrightEdge Generative Parser™ detects shopper offer cards which offer pricing, as well as offers such as free delivery in a compact view. While it remains uncertain if these are sourced from the Schema "offer" tag, it's evident that signals aligning with what the offer card displays could enhance the likelihood of appearing in these results:

Hero Images

Currently, when activated, these elements consistently appear at the bottom of the answer box. Notably, all images appear to be sourced. To enhance eligibility for display in these sections, it is crucial to ensure that images are optimized with relevant tags and are optimally compressed. This strategy underscores the importance of image optimization in facilitating effective and relevant search result displays:

YouTube Embeds

AI Overviews often display YouTube videos as citations for instructional queries. These videos frequently start at specific timestamps where the relevant content begins. BrightEdge Generative Parser™ has observed this timestamp functionality working with and without embedded transcripts, suggesting the AI truly processes video content directly. Marketers creating tutorial content for potential citations should host videos on YouTube, as this appears to be the only video source used.

How do AI Overviews (AIO) overlap with traditional results

AI Overviews are designed to provide additional context to queries that users may find helpful. As a result, many of the citations used by AI Overviews are different from those that rank in traditional results. This is because AI Overviews aim to answer the next question a user might have after their initial query, without prompting the user to ask that second question.

Across all keywords, we see that the URLs cited in AI Overviews are not ranking in traditional results 66% of the time.

The focus on content that isn’t currently ranking is a result of Google focusing on making AI Overviews as helpful as possible. Instead of simply repeating the information users could find within the rankings, the AI Overview anticipates the next question users would have and cites content that likely wouldn’t fit the immediate keyword ranking.

For marketers, this presents an opportunity to address multiple questions with multiple URLs in a single search result.

What Domains Are Sourced the Most?

As indicated, we are seeing our data skew to healthcare. Despite concerns about citation trustworthiness in the news, at the macro level, the share of citations supports the notion that the more trusted and authoritative the website is, the more likely it is to be cited.

Within each industry, the websites cited to support AI Overviews also vary. We also see that within industries, there are leaders that occupy a greater share of citations than others.

E-Commerce

Rank Overlap for E-Commerce

B2B Technology

B2B Rank Overlap

Education

Education Rank Overlap

Entertainment

Rank overlap for Entertainment

Finance

Rank overlap for Finance

Healthcare

Rank Overlap for Healthcare

Insurance

Rank Overlap for Insurance

How do AI Overviews (AIO) Differ from Search Generative Experiences (SGE)?

AI Overviews are the refinement of a series of tests and deployments that Google conducted for nearly a year with Search Generative Experiences. As a result of these tests, there are refined conditions that have strategic implications.

AI Overviews do not offer follow up questions

With SGE, Google prompted users to ask a followup which would initiate a conversation with the AI:

AI Overviews have scaled back this functionality, which may indicate that Google does not want to cannibalize engagement from the rest of the search result page.

Today, AI Overviews are appearing in significantly fewer queries, implying that Google has fine-tuned where AI Overviews are most helpful, thus reducing the need for follow-ups. Additionally, many anticipated questions are addressed directly in the AI Overview, negating the need for user prompts. For example, in the query “what are training shoes,” the AIO anticipates a question about the differences between training shoes and running shoes and answers it immediately, without requiring further interaction from the user. It also cites sources for answers to anticipated questions.

AI Overviews are 20% Smaller

Across the board, due to changes in how modules are presented and how content is consolidated, AI Overviews take up less space on the screen than SGE did. This means that other aspects of the search result page, such as traditional organic results and sponsored ads, are less likely to be crowded out when they appear. On average, the BrightEdge Generative Parser™ has detected that these are 20% smaller than their SGE counterparts. The difference in screen real estate is noticeable.

For marketers, this means that AI Overviews are less likely to diminish clicks from other elements of the search result page. However, it also means that citations and the opportunity to gain real estate within the AIO for a keyword are likely to be more competitive, as there is less space to compete in.

While SGE allowed users to respond with a vote on whether they liked a result or not, AI Overviews allow users to provide specific feedback on the result to help Google fine-tune it. This is a particularly important feature as the AI Overview rollout has faced some challenges with accuracy. Additionally, it gives marketers an opportunity to report inaccuracies on their own branded queries. For example, if a query contained inaccurate information about Macs, the digital marketer could provide that feedback directly to Google.

For marketers, this is an opportunity to include brand accuracy into workflows. Monitor brand queries for the presence of AI Overviews and use the Feedback Module to address any inaccuracies or editorial content that isn’t accurate.

AI Overviews are using more space efficient modules to display products

Ecommerce is a category where AI Overviews are appearing more frequently than on average (20% instead of 10%). Throughout the SGE testing period, the most common way to display products was through a module called a Product Viewer, which featured the product standalone with content generated by Google.

With the launch of AI overviews, these results were decline in visibility as they became more in line with carousel results:

We saw an overall decrease in the presence of product viewer formats when AI Overviews were launched, as Google reduced the overall deployment of AIO. Since that time, however, we are seeing a steady rise of carousel groups, which allow for more products to appear in less space than a product viewer.

For marketers in ecommerce, this means it is critical to ensure your product content has the appropriate schema markup and context, including reviews, images, pricing, and supplementary information such as inventory counts, so AI can present them alongside competing products.

Conclusion

A lot of the codebase and principles uncovered during the lab phase of SGE carry through to AI Overviews. However, with AI Overviews being more selective in where they appear and serving a more defined purpose, marketers need to ensure sensing and reacting to them is part of their content strategy.

With BrightEdge, marketers can leverage Data Cube X to filter on features such as Featured Snippets to pinpoint which of their keywords are likely to trigger an AI Overview. These are then integrated into a workflow to ensure they are optimized for both traditional rankings and AI Overview citations.

The data across these industries demonstrate how AI Overviews do not rely on the same ranking methods as traditional algorithms. In many cases, AI Overviews are sensing and anticipating questions that may also be asked. To build content that addresses both current and anticipated questions, marketers can leverage BrightEdge Copilot for Content Advisor. This tool uses AI to gather all the concepts and questions related to a topic to generate briefs and first drafts that will win in traditional search and be ideal for AI Overview citations.