AI model for Google search: Multitask Unified Model to improve search results


A new AI model, the so-called Multitask Unified Model (MUM), is intended to improve Google's search results in the future and thus emphasize sources that demonstrate authority, expertise and trustworthiness even more. The info boxes highlighted in the search result, the featured snippets, also benefit.

Google's systems recognize consensus thanks to an AI model

As the search engine giant Google announced on its in-house blog, the results of the search should be determined even more accurately and intelligently in the future. In the article “New ways we're helping you find high-quality information”, the search engine operator explains how he wants to achieve this goal.

A new AI model, the so-called Multitask Unified Model (MUM), which Google presented for the first time in May 2021 and called “A new AI milestone for understanding information” at the time, should help here.

The AI ​​helps Google's systems to understand the concept of consensus, for example when several sources on the Internet that are known to be reliable publish the same information and largely agree on the facts.

By using our latest AI model, Multitask Unified Model (MUM), our systems can now understand the notion of consensus, which is when multiple high-quality sources on the web all agree on the same fact.


The info boxes highlighted above the search results, the so-called featured snippets, which Google calls “highlighted snippets” in German, should also benefit from the new AI model. The featured snippets are a format that is intended to provide the user with a direct answer to his question, so that he does not have to visit a website at all. Demonstrated here using the search query for the topic “SLI”.

The featured snippets are intended to provide a quick answer to questions

The info boxes highlighted in this form in the search and integrated before the actual search results can be integrated into the search as text snippets, video snippets and list and table snippets.

Our systems can check snippet callouts (the word or words called out above the featured snippet in a larger font) against other high-quality sources on the web, to see if there's a general consensus for that callout, even if sources use different words or concepts to describe the same thing.


The Multitask Unified Model and other AI models also help Google's systems understand when featured snippets are not the right way to present information present and answer questions, explains the group.

We've found that this consensus-based technique has meaningfully improved the quality and helpfulness of featured snippet callouts.


Google provides more context to search results

A new “About the Source” page is also intended to provide further information on the websites and the companies behind them that serve as sources for the search results.

In the future, the search will provide further information on sources (image: Google)

In Germany, this feature will be called “About this result” and is intended to help users learn more about the context of a search result even before it visits a web page. For search queries for which very little information is currently available and for which the AI ​​cannot find a consensus, Google will in future issue a corresponding message and the recommendation to carry out the search query again later.

If there is little reliable information, Google will point this out in the future (Image: Google)

The The rollout of the Multitask Unified Model (MUM) and the new functions has already begun and, as usual, should gradually arrive in Germany in waves “in the next week”.