Google’s Query Fan-Out Patent and Thematic Search

A recently filed Google patent called Thematic Search appears to align closely with the Query Fan-Out method used in AI Mode. Submitted in December 2024, this patent sheds light on how Google’s AI systems generate responses to complex queries and also offers insight into future content strategy approaches.

The core idea behind Thematic Search is to improve how search results are organised for users. Instead of listing links individually, the system groups related results into broader categories—referred to as “themes.” Each theme is then summarised, offering users a quick understanding of different aspects of their query without needing to visit every website.

This approach is particularly designed for handling broad or in-depth research questions. What makes it innovative is how it uses AI to automatically detect patterns and topics from standard search results. It then generates brief but informative summaries for each theme, drawing on both the content and the context of the search listings.

In practice, this means users can access more structured and digestible information directly within the search results, especially when queries involve multiple layers or topics. For marketers and content creators, this signals a shift in how search engines present information and could reshape how content is planned and optimised in the future.

Thematic Search Engine

The concept of “themes” in search is not new — it dates back to the early days of search engines. That’s why this particular patent caught my attention a few months ago and prompted me to bookmark it.

At a glance, the patent outlines a system that integrates with both a traditional search engine and a large language model. It includes the use of a summary generator, which plays a key role in delivering more organised and useful information.

Essentially, when a user enters a search query, the thematic search engine processes the input and sends it to a standard search engine. Once results are returned, the thematic system takes over, sorting the results into relevant “themes.”

From there, the technology uses a large language model to generate concise summaries for each of these grouped themes. This makes it easier for users to digest a wide range of results without needing to click through every single link.

One of the more interesting aspects of the patent is how it handles broad search terms. A single user query can trigger several sub-queries, each centred around a particular sub-theme of the original topic.

In short, the patent describes a more structured way of presenting search results, one that could greatly enhance the user experience, especially for complex or multi-faceted questions.

Comparison Of Query Fan-Out And Thematic Search

The system outlined in the patent closely mirrors what Google describes as its Query Fan-Out technique in official documentation.

According to the patent, when a search query is submitted, the thematic search engine can generate thematic data from at least part of the search results. For example, it may take those results and identify more specific themes — or sub-themes — such as “neighbourhood A,” “neighbourhood B,” or “neighbourhood C,” based on the content of the documents retrieved.

These sub-themes can then be displayed on the search results page. If a user selects one of these, the process continues: the thematic search engine will fetch a new set of results from the search engine and break those down into even more specific groupings — essentially, sub-sub-themes — allowing for increasingly refined results.

Google’s own documentation explains the Query Fan-Out technique in a very similar way. It describes how multiple related searches are launched simultaneously, each exploring different subtopics across various sources. The system then pulls together these results to deliver a more comprehensive and easy-to-digest answer.

This layered method allows users to uncover a wider and deeper range of information compared to a standard search, making it easier to explore more nuanced or complex queries.

Summary Generator

The summary generator forms a crucial component of the thematic search system. Its primary function is to create brief written overviews for each theme identified from the search results.

In many cases, this generator is powered by a large language model designed to produce original content. It works by selecting specific passages from the search results that are associated with a particular theme.

To enhance the relevance and quality of the summaries, the system may also incorporate additional information. This includes page titles, metadata, and related content found near the selected passages, providing helpful context.

The summary generation process can be triggered in two ways. It may start when a user enters a search query, or when the thematic search engine is activated. The patent does not go into detail about what “initialisation” means, but it likely refers to the system getting ready to handle a search request.

Ultimately, the summary generator helps transform thematically grouped search data into clear, concise summaries. These summaries aim to present the most useful information based on what the user is looking for.

Query Results Are Clustered By Theme Instead Of Traditional Ranking

According to the patent, traditional search results can sometimes be replaced by grouped themes accompanied by auto-generated summaries. This thematic approach changes both the type of content shown and how it is presented to users. Instead of displaying a simple list of links, thematic search encourages users to explore deeper layers of information through related themes.

This shift could have a significant impact on publishers and SEO professionals. A typical keyword or phrase optimised for search may now only serve as the beginning of the user’s journey. As users explore the thematic structure, they may follow sub-themes that guide them to content slightly removed from their original query. Consequently, the website that gains the click may not be the one ranked first for the initial search term, but rather a different site that better fits a closely related or adjacent topic.

The patent outlines several ways the thematic search engine could function. These methods help organise and display information more intuitively:

Themes can be shown on the search results page, and in some cases, the results themselves are grouped or sorted based on these themes. The theme might be represented by a clear phrase visible to the user.
In some implementations, the thematic engine ranks the themes according to their relevance or prominence in relation to the user’s query.
The search results might be categorised under specific themes — for instance, results connected to “cost of living” would be grouped accordingly.
Presentation of these results may vary, including layout formats like lists, cards, or carousels (both horizontal and vertical), depending on the interface.
Thematic search results are sometimes shown alongside traditional search results, offering users a richer selection of related information.
In other cases, the theme-based content is displayed in a separate section of the page, distinct from the regular search engine listings.

This system allows for a more layered and exploratory search experience, potentially increasing the visibility of a wider range of content while changing how websites are discovered.

Thematic Search: Implications For Content & SEO

There appear to be two ways in which AI Mode may impact publishers at the conclusion of a user’s search journey.

Firstly, users might find the answers they need directly within the theme summaries or dropdown sections provided by the AI system. This could lead to an increase in zero-click searches, where users don’t need to visit a website at all, potentially reducing traffic from traditional search results.

Alternatively, the user might continue exploring until they reach a specific web page that effectively concludes their search journey. In this case, the site that provides the final, most relevant answer is the one that receives the click — even if it’s not the top-ranking result for the original query.

This shift suggests we may need to rethink the conventional approach to keyword rankings. Instead of focusing solely on ranking for a single term, it may be more useful to consider what question each page is answering. From there, it could be helpful to anticipate related or follow-up questions and either integrate that information into the existing content or create separate pages that address those additional topics — especially if they represent the final step in a user’s search journey.