In a recent interview, Google’s VP of Product for Search, Robby Stein, shared new information about how query fan-out works in AI Mode.
Although the existence of query fan-out has been previously detailed in Google’s blog posts, Stein’s comments expand on its mechanics and offer examples that clarify how it works in practice.
Background On Query Fan-Out Technique
When a person types a question into Google’s AI Mode, the system uses a large language model to interpret the query and then “fan out” multiple related searches.
These searches are issued to Google’s infrastructure and may include topics the user never explicitly mentioned.
Stein said during the interview:
“If you’re asking a question like things to do in Nashville with a group, it may think of a bunch of questions like great restaurants, great bars, things to do if you have kids, and it’ll start Googling basically.”
He described the system as using Google Search as a backend tool, executing multiple queries and combining the results into a single response with links.
This functionality is active in AI Mode, Deep Search, and some AI Overview experiences.
Scale And Scope
Stein said AI-powered search experiences, including query fan-out, now serve approximately 1.5 billion users each month. This includes both text-based and multimodal input.
The underlying data sources include traditional web results as well as real-time systems like Google’s Shopping Graph, which updates 2 billion times per hour.
He referred to Google Search as “the largest AI product in the world.”
Deep Search Behavior
In cases where Google’s systems determine a query requires deeper reasoning, a feature called Deep Search may be triggered.
Deep Search can issue dozens or even hundreds of background queries and may take several minutes to complete.
Stein described using it to research home safes, a purchase he said involved unfamiliar factors like fire resistance ratings and insurance implications.
He explained:
“It spent, I don’t know, like a few minutes looking up information and it gave me this incredible response. Here are how the ratings would work and here are specific safes you can consider and here’s links and reviews to click on to dig deeper.”
AI Mode’s Use Of Internal Tools
Stein mentioned that AI Mode has access to internal Google tools, such as Google Finance and other structured data systems.
For example, a stock comparison query might involve identifying relevant companies, pulling current market data, and generating a chart.
Similar processes apply to shopping, restaurant recommendations, and other query types that rely on real-time information.
Stein stated:
“We’ve integrated most of the real-time information systems that are within Google… So it can make Google Finance calls, for instance, flight data… movie information… There’s 50 billion products in the shopping catalog… updated I think 2 billion times every hour or so. So all that information is able to be used by these models now.”
Technical Similarities To Google’s Patent
Stein described a process similar to a Google patent from December about “thematic search.”
The patent outlines a system that creates sub-queries based on inferred themes, groups results by topic, and generates summaries using a language model. Each theme can link to source pages, but summaries are compiled from multiple documents.
This approach differs from traditional search ranking by organizing content around inferred topics rather than specific keywords. While the patent doesn’t confirm implementation, it closely matches Stein’s description of how AI Mode functions.
Looking Ahead
With Google explaining how AI Mode generates its own searches, the boundaries of what counts as a “query” are starting to blur.
This creates challenges not just for optimization, but for attribution and measurement.
As search behavior becomes more fragmented and AI-driven, marketers may need to focus less on ranking for individual terms and more on being included in the broader context AI pulls from.
Listen to the full interview below:
Featured Image: Screenshot from youtube.com/@GoogleDevelopers, July 2025.