Google MUM Update of SEO 2022 & MUM vs BERT

Last May, Google has rocked the world with the launch of key MUM Algorithm. Want to learn what’s exactly MUM is, what impacts are for your SEO and MUM vs BERT comparison, please read on.

What is Google MUM exactly?

The Google MUM (stands for Multitask Unified Model) , released in May 2021 during the Google IO Conference, is one of Google’s biggest algorithm updates this year. It has activated in an effort to improve search results, and users have already experienced improvements in the way they conduct searches for answers.

According to Google’s blog post, MUM algorithm will be exploring complex tasks with both text and images, which quickly turn on the alarm of SEOers that call for immediate attention. Old way of doing SEO only requires us to take care of the 10 blue links each page of SERP, however, it’s no longer valid with the single dimension SEO.

MUM is 1000 times more powerful than BERT

MUM stands for “Multitask Unified Model”

MUM is 1000 times greater than BERT

According to Google’s blog post, it deploys the newer transformer in MUM, which is 1000x more powerful than BERT. It can handle multiple tasks at the same time. In the I/O event, it showed the power of unripping all context behind the topic (a complex question related to Mt. Fuji was the example), taking reference from multiple sources and feedback with a more complex answer. Google takes one giant step further to handle complex queries with the following key areas:

  1. Extracting information from multiple sources, at the same time
  2. Breaking the barrier of language (potentially the culture barrier by following direct source of information)
  3. Incorporate video and images, then presented in a rich snippet

It also supports Google LENS feature then turn image processing into answers. So, it’s no longer a sci-fi that a super human brain answering everything. AI human-like search experience is in your sight.

Impactful SEO Campaign in 2022 onwards

Think content from your audience perspective would be rule no.1 of your ongoing SEO strategy. Google’s MUM update forces you to shift the thin content strategy into a more grand view of considering topic and all subtopic which could assist your audience to find the answer. Apart from it, you should now consider using rich content (video, image or sounds) in your content to support your reach. You shouldn’t just stop there, sub-sequential content should be planned in pipeline if you want to own the voice of such topic.

Best Approach for SEO 2022

It’s the best timing to change your SEO strategy 2022 because Google has already started testing MUM algorithm in popular topics. You could apply the new methodology and see what impacts the most in ranking and traffic in your own keyword category.

Here’s my recommended high level to-do list:

  1. Change keyword research to topic research now. If you still consider each keyword as individual result, you might lose the ranking if someone else provide a more intelligent answers to your keyword.
  2. Content is King vs Quality is King. Content remains a key elements in making your website rock, yet quality convey your website to a more sustainable growth. Your task becomes more challenging with MUM being introduced. Because not only engaging content works better, but also shows dominance of topic works even better.
  3. In 2019 with BERT update, I would tell you that you should provide answers to search queries. Now, entering 2022, I would say that providing solutions to search queries is a better idea.
  4. “Intuition is a very powerful thing, more powerful than intellect, in my opinion,” Steve Jobs once said. Following data is a good way to explore the topic, but sometimes intuition should kick in when you comes to providing solution to your audience, because no one knows your audience better than you do.
Key Reference Links:

Official Blog: https://blog.google/products/search/introducing-mum/

Exploring Transfer Learning with T5: the Text-To-Text Transfer Transformer: https://ai.googleblog.com/2020/02/exploring-transfer-learning-with-t5.html

HyperGrid Transformers: Towards A Single Model for Multiple Tasks: https://research.google/pubs/pub49971/