August Brings Significant Changes in AI Overviews Rollout and Behavior

August was a busy month for Google AI Overviews (AIOs), bringing significant changes in how AI-generated results are rolled out for searches across different industries and user types. From variations between logged-in and non-logged-in users to shifts in rank overlap and product viewer trends, we observed fascinating new developments that indicate Google is continuing to refine its AI-driven search capabilities.

Here are the top 7 takeaways from what we observed in August.

1. The AI Overviews rollout to non-logged-in users is not evenly distributed across industries.

One of the most striking changes in August was the rollout of AI Overviews to non-logged-in users, which began around the middle of the month. This marks the second phase of the AI Overview rollout.  The first phase began around May 15th when AIOs started showing to logged in users who had not registered to Google Labs during the previous testing phase.  Between now and then we’ve witnessed numerous adjustments and fine tuning to results which we’ve tracked in our Ultimate Guide to AI Overviews. With this second phase where AI Overviews now appear for non-logged in users, we can finally consider AI Overviews officially live.   

BrightEdge Generative Parser™ shows that while some industries experienced relatively consistent integration of AI Overviews between logged-in and non-logged-in users, others, particularly eCommerce, saw significant discrepancies. Overall, the data suggests that Google is rolling out AI Overviews with fewer keywords for non-logged-in users during these early weeks. The main points we observe in these first couple weeks of rollout are:

• AI Overviews are less visible on average (~10%) for non-logged in users- this suggests Google is being more cautious on where to deploy them. It could also be that for non-logged in users, Google is operating on less insights into the users’ preferences and thus, only deploying AI Overviews when the query intent is very clear. 
• For Ecommerce in particular, the difference between logged in and non logged in users is significant, with only 1% of keywords showing an AI Overview. There are a few reasons this could be happening which I’ll layout below.
  
•  B2B Tech: AI Overview presence for logged-in users was 5 percentage points higher compared to non-logged-in users, resulting in a 17% relative drop.
• eCommerce: The most significant gap was observed here, with non-logged-in users seeing 90% fewer AI Overviews compared to logged-in users. This suggests that Google is taking a much slower approach to rolling out AI Overviews in eCommerce.
• Education: There was a 9 percentage point drop in AI Overviews for non-logged-in users, translating to a 21% relative change.
• Healthcare: AI Overview presence was 10 percentage points lower for non-logged-in users compared to logged-in users, resulting in a 16% relative change. This smaller difference suggests a more deliberate rollout in the healthcare sector.

  Overall, the data shows that AI Overviews are rolling out unevenly across industries, with eCommerce queries experiencing the slowest adoption among non-logged-in users. The complexity of product searches and the need for more refined decision-making frameworks likely contribute to this slower rollout.

Why the Slow Rollout for eCommerce?

The massive 90% drop in AI Overview presence for eCommerce queries among non-logged-in users is striking. This could be due to the complex nature of eCommerce searches, where users are often comparing multiple products, prices, and reviews. Google may be taking extra care to ensure that AI Overviews in this space provide accurate and useful information before fully rolling them out to non-logged-in users. With e-commerce one of the largest industries for Google Ad Revenue, it makes sense they would be cautious with anything that could potentially alter the search result page.

Another possibility is the evolving role of product viewers in regular search results. BrightEdge Data Cube X data shows significant increases in the product grid, which takes up substantial screen real estate for non-logged-in users and is highly transactional. It may be that, for non-logged-in users, Google assumes a more transactional intent compared to logged-in users, where they may have more data on individual user behavior. The chart below shows the top share of clicks across eCommerce queries tracked via Data Cube X:

An example of how this discrepancy between a logged in and non-logged in search experience appears is for the term “Chinos”. When logged in, an AI Overview appears with an apparel product carousel along with fashion guidance on what wear with chinos :

When this query is done without logging in, there is no AI Overview but the product grid now occupies a significant portion of the search result page:

As Google gathers more data on how to effectively deploy AI Overviews for non-logged-in users (i.e., without historical knowledge of the user), we may see them being rolled out more frequently.

2. Rank overlap fluctuations correspond to Google’s August core update.

Google’s August 15th Core Update has introduced noticeable shifts in how AI Overviews overlap with regular search rankings. Google’s John Mueller confirmed that these core updates will impact AI Overviews, as they are part of search (although he did not specify how exactly). BrightEdge Generative Parser™ reveals that the overlap of URLs cited in AI Overviews with those ranking in the top 100 increased from 37% to 41% post-update. Notably, the data now indicates that Google is incorporating more organic results into these citations, specifically pulling from lower-ranked results in the SERPs to leverage relevant but less prominent links more effectively.

This insight also suggests AI Overviews do have some nuances that separate them from organic results. Contrary to claims that optimizing for organic search would suffice for inclusion in AI Overviews, our data does not support this conclusion. This misunderstanding could stem from an oversimplification of how AI Overviews are generated. AI Overviews do not solely rely on top organic rankings; instead, they involve a more complex selection process that considers a variety of factors beyond standard SEO practices. This includes the relevance and authority of the content, what the AI anticipates will be the second and third query to address, and its overall utility in a given context, which may not always align with traditional SEO metrics.

This increase in rank overlap suggests that Google is prioritizing content already performing well in traditional search rankings for inclusion in AI Overviews. However, the average rank of URLs in both AI Overviews and regular results has dropped slightly—from 10 before the update to 11 after—indicating that lower-ranking pages may now be featured more prominently in follow-up AI Overviews.

This change also impacts different industries in distinct ways:

• B2B Tech and Entertainment saw improvements in rank overlap, with B2B Tech climbing from an average rank of 6 positions, suggesting these sectors are benefiting from the update. For example, for the keyword ‘access control entry’ the domain calcomsoftware.com is cited in the AIO and ranking in the 6th position. In entertainment, for the query ‘mysteries’ the domain history.com is cited in the AIO and ranking in the 6th/7th position.
• Education and eCommerce, on the other hand, experienced declines, with Education dropping 12 positions on average and eCommerce slipping 3 showing that these sectors may need to focus more on optimizing for follow-up queries or AI-driven referrals rather than traditional rankings alone. In ecommerce, an example of this movement is with the query ‘October birth stone’ in which the domain jewelryofstuart.com which ranks in the 12th position is now cited in the AIO.  For Education, the query “Cyber security salary” tends to draw on data from sites such as Glassdoor which ranks on the second page on Google.

For marketers, this means that lower-ranking content could still gain visibility if it addresses user queries following the initial search. Optimizing your pages for related and tangential questions will be key to capturing traffic from AI Overviews in the post-update landscape. This performance boost also highlights how core updates can deliver incremental value beyond rankings. As the overall quality of the index improves, the chances of deeper-ranking content being sourced by AI Overviews increases. Ensuring that your content follows Google’s guidelines not only enhances your traditional organic ranking opportunities but also increases the likelihood of being cited in AI Overviews.

3. AI Overviews are evolving in how they present information.

August brought some noticeable shifts, with changes aimed at presenting information more efficiently and in a user-friendly manner. Our data at BrightEdge reveals a series of significant adjustments between July and August, particularly relevant for marketers optimizing content related to products.

1. More Efficient Information Presentation
   We observed a 172% increase in product carousels for apparel-related queries. This suggests that, rather than focusing on a single product, AI is increasingly providing users with broader choices and comparisons, offering a more comprehensive view of available options.
   
    
2. Smarter Content Organization for eCommerce Products
   We also observed a 42% increase in the use of unordered lists across industries. In the eCommerce space, this means AI Overviews are better organizing product features and specifications, making it easier for users to scan through options and make informed decisions quickly.

   In addition, comparative content is becoming more prevalent, with a 12% rise in product comparisons in AI Overviews. This shift indicates that products are being evaluated not just on their individual merits but also in relation to similar items, allowing users to make side-by-side comparisons more easily. These include complex queries and products that may require more research before making a decision.
   

These changes reflect a broader trend toward more comprehensive, comparison-based content in AI Overviews. By adapting your strategy to these evolutions, you can enhance your content’s visibility and engagement in an increasingly AI-driven search environment.

4. AI Overviews are exposing more citations

On August 20th, AI Overviews changed how they displayed citations.

Historically, Google’s AI-powered results provided citations through a chevron within the answer as well as a citation box to the right of the result:

Potentially in response to the launch of SearchGPT and its expanded citation panel, Google has updated its citation boxes to make it easier for users to explore more content related to a topic. Now, the citation box looks like this:

Interestingly, these new sources do not increase the size of the AI Overview panel on their own, as the box is scrollable, making it easy to read both the AI Overview response and browse for more information. Additionally, users can now click on the link icon to get a specific view of the citations for that section of the answer. This is particularly useful for informational queries, such as ‘lupus,’ where the intent may be ambiguous. Now, if users find exactly what they want to learn more about, they can pinpoint it directly in the result and dig deeper into the subject.

For marketers, this means there is an even greater chance for content to drive referrals for long-tail queries that users may not have originally searched for, as the AI Overview attempts to surface deeper content on broad topics like this one.

5. After months of shrinking, AIO are getting bigger

One of the most notable changes observed in August is the significant increase in the height of AI Overviews. This is important because, for several months, BrightEdge Generative Parser™ had detected a continuous reduction in the size of the AI Overview panel. However, the average height of AI Overviews rose by 10%, from 647 pixels in July to 712 pixels around August 18th. This increase may seem subtle at first glance, but it has profound implications for how AI Overviews present information and what it means for marketers looking to optimize their content for visibility. Additionally, this height increase coincides with the rollout of AI Overviews to non-logged-in users.
What Does a Larger AI Overview Panel Mean?

The increase in AI Overview height suggests that Google is making more room for rich, detailed content within the AI Overview itself. Here’s what this trend tells us:

1. More Room for Comparisons and Lists
With additional space, AI Overviews can now accommodate more comprehensive, comparison-based content. This aligns with the rise in unordered lists and product carousels we’ve observed, particularly in industries like eCommerce. The added space allows Google to present products side-by-side, including more product features, comparisons, and visuals that help users make better decisions more quickly.

2. Increased Visual Content
The extra real estate is also being used to incorporate more visual elements. The trend toward larger, richer images—especially for product-related searches—is becoming increasingly important. For eCommerce marketers, this means ensuring that product images are optimized not just for quality but for visibility within AI Overviews’ product carousels. High-quality images are essential to drive clicks, particularly when they are used in carousels alongside other products.

3. Deeper, More Nuanced Information
Larger AI Overviews also indicate a shift toward more detailed, nuanced information. Instead of simply providing a brief answer or product summary, AI Overviews are now offering users expanded explanations, deeper insights, and additional context. This benefits users by allowing them to access more comprehensive answers directly within the search results, without needing to click through multiple pages.

6. Domain Citation Changes: Shift Toward Specialized Sources

BrightEdge Generative Parser™ data reveals some interesting fluctuations in domain citations between July and August. While certain domains saw notable increases in citations, others experienced reductions, indicating a preference for more specialized or authoritative sources in specific fields.

1. Medical and Health Domains
   There was a significant shift in health-related citations, moving away from general health information sources towards more authoritative and specialized medical sites. General health platforms like verywellhealth.com and everydayhealth.com saw substantial decreases in citations (-77.9% and -95.6% respectively). In contrast, authoritative medical sites such as mayoclinic.org (+32.4%) and nih.gov (+83.2%) experienced notable increases. Even more striking was the surge in citations for specialty health sites, with spine-health.com seeing a 266.7% increase and arthritis.org an 89.5% rise. This pattern strongly suggests a focus on higher authority, more specialized health information in AI Overviews, with a growing preference for both comprehensive medical resources and niche sites providing in-depth content on specific conditions or treatments.
2. Government Domains
   While nih.gov saw a significant increase in citations (+83.2%), other government domains like fda.gov saw a slight decline (-10%). This suggests selective use of government sources, with an emphasis on health-related institutions.
3. News and Media Outlets
   There was a 3% decrease in citations for usatoday.com and a 3.1% decrease for nytimes.com, indicating a possible shift away from general news outlets. In contrast, specialized news sources like bloomberg.com saw an increase in citations, reflecting a preference for more focused, industry-specific information sources.
4. Technology and Product Review Sites
   Tech review sites like techradar.com (-47.3%) and tomsguide.com (-16.4%) saw declines in citations. This aligns with the overall trend of AI Overviews relying less on consumer technology information and more on industry-specific or expert sources.

These changes suggest that AI Overviews are becoming more selective, favoring highly specialized, authoritative content over general sources. In particular, the healthcare and educational sectors saw shifts toward more trusted, expert-backed information.

7. Keyword Presence: Emphasis on Specific and Technical Content

BrightEdge Generative Parser™ data reveals significant changes in AI Overview coverage based on keyword characteristics between July and August. The analysis shows a clear trend to show AI Overviews on more specialized, technical, and factual keywords across various subjects.

1. Medical and Health Keywords
   There was a substantial increase in AI Overview coverage for specific medical conditions and symptoms (+118.0%). For example, the keyword “acl tear symptoms” gained AI Overview coverage, reflecting the trend toward more comprehensive, expert-backed health information in AI Overviews.
2. Technical and B2B Concepts
   Keywords related to technical concepts, security, and data processing in the B2B tech sector saw a significant increase in AI Overview coverage (+71.4%). An example is “what is continuous integration,” demonstrating the growing emphasis on providing AI-generated overviews for complex, industry-specific topics.
3. Product Specifications in eCommerce
   Keywords focusing on product features and specifications experienced the highest percentage increase in AI Overview coverage (+125.7%). For instance, “tv stand with fireplace” gained AI Overview coverage, indicating a strategic focus on detailed, factual product information rather than broader, less specific product descriptions.
4. Financial and Stock-related Terms
   Finance keywords, particularly those related to stocks and financial metrics, doubled their AI Overview coverage (+101.4%). An example is “wtrh stock,” suggesting a careful expansion into more specialized financial topics.
5. Entertainment and Travel Topics
   In contrast to the increases in technical areas, entertainment, travel, and restaurant keywords saw a complete removal of AI Overview coverage (-100%). For example, “things to do in Sacramento” lost its AI Overview coverage, indicating a move away from more subjective or rapidly changing content in favor of more stable, factual information.
6. Specificity Across Domains
   Across all categories, specific or technical keywords (e.g., “data aggregation”) saw a substantial increase in AI Overview coverage (+77.7%), while broader or general interest topics (e.g., “restaurants near me”) experienced only minimal growth (+11.3%). This disparity underscores a clear strategy to prioritize AI-generated content for more complex, specialized queries that benefit from comprehensive, objective overviews.

These changes in keyword characteristics mirror the changes observed in domain citations, pointing to a cohesive strategy in AI Overview generation. The focus continues toward providing in-depth, authoritative information on specific topics, particularly in fields like healthcare, technology, and finance. This trend suggests that AI Overviews are being optimized to serve as reliable sources of specialized knowledge, catering to users seeking detailed, expert-level information rather than general overviews or subjective content.

We are at a crucial inflection point with AI search. Not only has OpenAI entered the marketplace, but AI Overviews are also expanding their presence beyond logged-in users! As we monitor how this transition is unfolding, it’s fascinating to see how AI deployments are refining their focus to be most effective for users. For marketers, these advancements present amazing opportunities as the worlds of search and AI continue to converge!