RankBrain, Bert, Core Updates… Google updates regularly disrupt search engine results. The new algorithm called MUM could also fundamentally change the SERPs for years to come.
Always with the goal of giving the user the most relevant answer as quickly as possible, Google created MUM. This new algorithm, designed to respond to complex user queries, is 1,000 times more powerful than Bert, according to Pandu Nayak, Google’s vice president in charge of search engines.
Note that MUM uses T5’s text-to-text framework. As a result, the tasks of natural language processing, which includes understanding texts, classifying subject areas or even translating, can be processed more effectively. In other words: MUM understands the language and also generates it.
To illustrate MUM’s contribution, Pandu Nayak takes the example of a person who has hiked Mount Adams and wants to know what to prepare for a hike up Mount Fuji. As it stands, he would need at least eight queries to answer his question. MUM can process the complex query faster.
In the hiking example, MUM can understand that the intent of the query is a comparison between two mountains and consider information such as elevation and hiking trails to provide an answer. MUM can amplify their response to physical training or finding the right equipment. It’s actually multitasking. This is not new to Google’s algorithms, but MUM reinforces this character.
MUM will also be able to understand and display information from Japanese sites on Mount Fuji. Because MUM is multilingual. It can interpret and compare more than 75 different languages to show the result. “This allows Google to get answers from anywhere in the world, regardless of native language,” notes Mike Blumenthal, co-founder of Nearmedia.co.
MUM will also be able to respond to a user’s text request, which comes from a photo of their hiking boots. MUM is indeed multimodal. It understands the information provided by text and images, and will include video and audio in the future.
New features related to Google MUM
The features associated with MUM therefore include the ability to ask a question using a photo, but also to trigger visual SERPs on specific queries, to decide which images to extract from the selected pages and to invoke an advanced system of video recommendations display a “Fun things to know” feature that is more detailed than the current “People also ask”, or an “expand/refine this request” insert.
For some search queries, the results should therefore be even more appealing and sometimes require no clicks at all. Will all of this lead to a shift from a ranking-style SERP with ten blue links to a wiki-style SERP? “This process has already begun,” analyzes Florian Caux, SEO consultant at Procab Studio. “This is why sitelinks or People Also Asks show up. The likely goal is to replace websites with useful information that will appear directly in search engine results in the medium or long term. sees through airline tickets or hotel reservations. The fewer the fewer users go to the pages, the better Google performs.”
Which industries will MUM affect? “I think the long tail of every industry and every segment will be hit by MUM,” says Mike Blumenthal. “The goal of the algorithm is to better understand complex queries and query paths. Hopefully this will allow a user to get an answer by asking one challenging question instead of eight simpler questions.”
How can you work on your SEO with Google MUM?
While it’s difficult to know exactly what websites and content MUM is promoting, it may be wise to prepare for its arrival by optimizing certain SEO-specific elements.
Expertise, authority and trust
MUM requires a lot of processing power to display snippets. So you can imagine that Google first selects pages based on their relevance, quality and authority before extracting specific content elements on those pages.
It therefore seems prudent to start by respecting Google’s EAT, which encourages expertise, authority and trust. This is to be selected by Google before considering distributing excerpts of content from MUM. This is, quite classically in SEO, about showing authority with the right backlinks, bringing quality to the topic by integrating the core web vitals and expertise, for example by highlighting the profiles of specialist authors in your company.
“things, not threads”
The company Mountain View increasingly highlights relevant parts of a page in its results. This has spread to videos and should be confirmed with MUM. To help Google find relevant content on your pages, it can be a good idea to properly tag your content.
In addition to the “classic” HTML markup, you can also add markup from Schema.org. They allow Google to better understand the information and context of the page and help to better present the content in its results. Note that the Schema.org vocabulary can be used with many different encodings, such as B. RDFa, Microdata and JSON-LD.
What’s the point of markup with MUM? Markup is used to create entities and the relationships between them. For example, with schema.org it is possible to map a person to an organization and add a “same as” property. This helps mention links to authority sites and social media that include the person’s name.
“Schema.org allows, for example, a webpage’s code to indicate who the author is, whether they have articles on this or that authority site, or even a Wikipedia file,” notes Florian Caux. “It gives confidence to the site from the moment the author himself has confidence on the internet. It can be interesting to play for your authority.” One way to apply the “things, not strings” (things, not words) principle mentioned by Google ten years ago.
You can also use a markup generator like Google Structured Data Markup Helper to help you. You can choose the URL of the website and add elements like the author of the article, the publication date, the image, the title of the article…
cover user intent
In order to be able to use the “Worth knowing” and “Extend/refine search” tabs, you have to answer a large number of questions relating to your topic. An idea might be to work on your topic clustering to show Google your full approach to the topic on your site.
With this method, a pillar page ranks in the SERPs. Secondary pages evoke different knowledge about the topic. The meshing between the pages must be done with care. For example, by placing the link from a secondary page to the target page at the top of the page, or by placing the link from a secondary page to a final page supporting the secondary page at the end of the text…
Longer alt tags?
Alternative text improves accessibility by displaying text without an image. It can be interesting to specify this attribute to better contextualize the image. “The more meaningful the alt tag is, the better Google can understand the image and train its algorithm,” says Florian Caux. Indeed, in its best practices, the American multinational reveals “using not only the content of the page, but also the alternative text using computer vision algorithms to determine the object of the image”.
So far, a few descriptive words and the keyword were enough to fill out this tag correctly. With the arrival of MUM, the alt of images could expand. For example, in a recent article by Pandu Nayak on using AI in Google Search, the illustrative image had a very descriptive and long alt tag of… 38 words. A hint for the future?
An interesting method if this is the case, especially since Google wants to optimize multi-search with MUM. Remember, Google’s multi-search allows you to add a search to a screenshot or photo on Android and iOS. A very interesting application, especially for e-commerce inquiries…