Google’s MUM, the mother of all search algorithms?

Google’s MUM is changing ‘Search’ by overcoming language barriers, across diverse media types. Is MUM here to enhance SEO or render it obsolete?

In 2021, Prabhakar Raghavan, Senior Vice President at Google, announced the launch of a new AI model called Multitask Unified Model (MUM) at the Google I/O event. While the new model runs on the T5 framework, which is similar to BERT (Bidirectional Encoder Representation Transformer), MUM is superior to BERT by 1000 times. Google’s MUM is changing the way the search engine works as it goes beyond the traditional keyword search and cultivates a more comprehensive understanding of the user’s query. While there are many papers on the multi-model technology, one paper: “Rethinking Search: Making Domain Experts out of Dilettantes“, demonstrates the unified model approach, which is different from traditional IR systems as the model retrieves and ranks queries within the same component. 

Comparison between search models

Image: Comparison of traditional and unified models from Cornell University

According to Google, MUM can be applied in various tasks such as document summarising, sentiment analysis, question answering, etc. Such applications have made MUM a major priority at Googleplex headquarters. With MUM gaining importance in Google’s ecosystem, will it become a key factor in search rankings? Or will it render search engine optimization (SEO) obsolete?

MUM – Same brain, new synapse

Pandu Nayak, a Google Fellow and Vice President of Search, explains the problem statement: having to type out too many queries and perform various searches to get the answer you need. MUM is an update on the current search engine (BERT) as it is built on a text-to-text transfer transformer, one of the popular NLP models like LaMDA or GPT-3. 

MUM’s skillset can be classified into the following factors:

  1. Multimodal – MUM exhibits its capability to manage data in multiple formats. It can parse text from images and videos, giving it a multi-sensory experience. 
  2. Multitasking – MUM has the potential to bring a change in how Google helps users with complex tasks. As Pandu Nayak explains with an example, if the user asks a question about how to prepare for a hike to Mt. Fuji, MUM could not only provide the results for the particular query but go beyond and include things like fitness regimens and finding the right equipment.
  3. Multilingual – MUM is trained in over 75 different languages so that it can overcome language barriers and provide results even if the query is asked in a different language.

The debate – Will MUM affect Search?

SEO specialists and company brands constantly adapt to the ever-changing algorithms of the search engine. Google has been striving to bring a more natural experience to its search engine since its conception. Google’s MUM has the potential to reduce the time and effort invested by the user in complex searches. Gregg Turner, Head of performance marketing at UNRVLD, talks about the impact of MUM on SEO as he hints at the implications of MUM being able to pull relevant audio, video and image content, being huge to content businesses, agencies and stand-alone creators.

However, some experts are inclined toward the notion that MUM is an indicator of the shift in the way Search is perceived. Edwin Toonen, a strategic content specialist at Yoast, indicates in his post that the Google search engine is metamorphosizing into a “knowledge presentation machine”. Moreover, Google’s Vice President of Search acknowledges that the technology isn’t in play at present. When asked whether Google will announce when MUM will go live on Search, Google Search Liaison Danny Sullivan states that they won’t stay mum on MUM.

Twitter feed on Google's MUM

Source: Twitter


The power of Google’s MUM lies in its multimodal, multitasking and multilingual nature. It is poised to revolutionize the search industry, incrementing exponentially on the impact created by its predecessor BERT in 2019. While Google is not using MUM as a search ranking signal, it will train MUM on large datasets and fine-tune it for distinct applications on smaller datasets. This process is made evident by Google testing the model to improve vaccine search results. In this experiment, MUM proved efficient in identifying 800 variations of vaccine names in over 50 languages in just a few seconds.

In the present scenario, SEO professionals will have to be in a constant state of flux in order to adapt to the ever-changing search engine algorithms. There’s a possibility for a future where search engines are evolved enough to experience data as the user does, thus bridging the gap between query formats and results.

Download our Mobile App

Subscribe to our newsletter

Join our editors every weekday evening as they steer you through the most significant news of the day.
Your newsletter subscriptions are subject to AIM Privacy Policy and Terms and Conditions.

Our Recent Stories

Our Upcoming Events

3 Ways to Join our Community

Telegram group

Discover special offers, top stories, upcoming events, and more.

Discord Server

Stay Connected with a larger ecosystem of data science and ML Professionals

Subscribe to our Daily newsletter

Get our daily awesome stories & videos in your inbox