Google AI Overviews: New Research Offers Insights

New research by BrightEdge offers a snapshot of the kinds of queries that tend to show Google AI Overviews (AIO) and provides insights into the kinds of queries and verticals where AIO are more prevalent.

The findings show dramatic differences in the amount of AI Overviews shown across different verticals in a way that reflects the kinds of queries that are common. This effect works in reverse as well, where some verticals experience less AIO search features.

Is This A Paradigm Shift?

While BrightEdge calls it the greatest paradigm shift in decades, I think that’s understating shifts to Google search in the recent past, not just the ones in 2024. Something that’s not widely understood is that Google Search has been an AI Search engine since at least 2015 with the introduction of RankBrain and other subsequent changes to the backend side of search.

The big change in Search this year is that AI is more obvious on the front-end as a Feature in Search, largely replacing the role that Featured Snippets once played. Perhaps more importantly there may have been an infrastructure change at the beginning of 2024.

BrightEdge Generative Parser

BrightEdge has a technology, called the Generative Parser, which tracks and analyzes patterns in Google’s AI search features. BrightEdge used their Generative Parser to produce research findings about Google’s new AI Overviews (AIO) search feature.

Albert Gouyet, VP of Operations at BrightEdge said this about the BrightEdge Generative Parser:

“It’s fascinating to see the BrightEdge Generative Parser™ giving marketers a front-row seat into how AI in search is developing and giving the community a glimpse into the future. For marketers who rely on organic traffic, early indications suggest that AI will help reach new customers and present new opportunities to create content that serves multiple needs and elevates brand performance.”

What Triggers AIO

BrightEdge’s report indicates that Featured Snippets and questions were likely to trigger the AIO feature. Featured Snippets are answers to questions that are created with direct quotes from websites. BrightEdge found that AI Overviews were more likely to appear when there was also a Featured Snippet.

What Doesn’t Trigger AI Overviews

The research showed that local search queries were the least likeliest to trigger an AI Overview search result. That makes sense because a user is looking for a structured search result (business names, addresses, phone numbers), information that can’t be usefully summarized.

Similarly, search queries that generate sitelinks were also less likely to trigger AIO. Sitelinks are search results related to branded searches which feature multiple links to inner pages of a website. For example, searching for the name of a clothing store can generate a search result that features inner pages for women’s clothes, men’s clothes, etc. This also makes sense because it’s the kind of search query that is best answered with direct data and not a summary.

Verticals Most Likely To Contain AIO

Search results that tended to feature AI Overviews were wildly different when compared by verticals (verticals means specific industries or topics). This likely doesn’t mean that Google was targeting specific verticals for showing more AIO. Search features are always tied to the helpfulness of the features. The helpfulness of features are tested with the Search Quality Raters, workers who test out new kinds of search results and rate them for helpfulness and other criteria.

Search queries related to Healthcare tended to generate AI Overviews at a rate of 63% of the time. That makes sense for search queries that are information-seeking.

B2B technology queries tended to generate AIO results 32% of the time while Ecommerce search queries triggered AI Overviews 23% of the time.

Interestingly, restaurants and travel related queries did not tend to trigger AIO results.

AIO Shown Less Often Than SGE

Another interesting data point is that AIO is triggered 20% less times than Search Generative Experience (SGE) answers were.

BrightEdge offered three insights related to why AIO is shown less than the experimental SGE was.

1. “This indicates that AI is getting more precise when generating helpful experiences.
2. This is likely because AI now caters better to people’s needs, such as looking for summaries, recommendations, or conversational experiences.
3. Ultimately, Google is getting better at selecting answers.”

BrightEdge research pointed out that Google is improving the ability to anticipate follow up questions by providing AI search summaries that more completely answer a question.

They write:

“Since Google l/O, the overlap between citations in AI and traditional results has diminished. Google is ensuring users do not get the same results in the two types of different results. It is also now delivering on its promise to do the second, third, and fourth search for you. AI is beginning to anticipate the following question and give options before a user even asks. This often happens with ‘what,’ ‘where,’ and ‘how’ intent-based queries.”

Early Days Of AIO

Google has received overwhelmingly negative reviews from users and the news media about the quality of Google’s AI Overviews, which in turn can lead to trust issues. BrightEdge’s report can be considered a snapshot of Google AIO today and I’m certain BrightEdge will be back with new data in the future when Google’s (AI) SERPs eventually change again.

Featured image by Shutterstock/Marco Lazzarini