
Key takeaways
- Google announced MUM on 18 May 2021 as a new technology, with features promised over months and years.
- Google described MUM as 1,000 times more powerful than BERT and trained across 75 languages.
- Its first Search use identified over 800 vaccine name variations in more than 50 languages.
- Google's ranking systems guide says MUM is not currently used for general ranking in Search.
- There is no MUM-specific optimization; Google's helpful content guidance applies as usual.
What Google changed
MUM was not a ranking update. Google announced the Multitask Unified Model on 18 May 2021 as a technology it would bring to Search over time: “We’ll bring MUM-powered features and improvements to our products in the coming months and years.” There was no rollout window and no ranking change tied to the announcement.
In that post, Google’s Pandu Nayak said MUM uses the T5 text-to-text framework and “is 1,000 times more powerful than BERT.” Google wrote that MUM both understands and generates language, is trained across 75 languages and many tasks at once, and is multimodal, handling text and images.
When it rolled out
The announcement date, 18 May 2021, is the date on this page because it is the date Google first described MUM. Google described MUM’s first application in Search on 29 June 2021. That was a specific task, not a general ranking change: identifying the names people use for COVID-19 vaccines, so Search could show trusted information for all of them.
What is known about its impact
Google gave one figure, for that first task: “With MUM, we were able to identify over 800 variations of vaccine names in more than 50 languages in a matter of seconds.” Google said that work would otherwise take hundreds of human hours. It published no figures on ranking or traffic, because MUM was not applied to general ranking.
What site owners should do
Google’s current ranking systems guide lists MUM as an AI system “capable of both understanding and generating language,” and says it is “not currently used for general ranking in Search.” It names specific uses instead, such as vaccine searches and featured snippet callouts. MUM is still listed as an active system, not a retired one.
- There is nothing to optimize for MUM directly.
- Write for people first, as Google’s helpful content guidance describes.
- Cover a topic in the language your audience searches in, clearly and completely.
How to check whether it affected you
MUM has no ranking rollout to measure against, and Search Console keeps only 16 months of data in any case. If you are reviewing traffic from mid-2021, attribute changes to the dated updates around it instead, such as the June 2021 and July 2021 core updates. The Google Algorithm Updates timeline lists them with their dates.
Related updates
- BERT (2019) was the language model Google compared MUM against.
- Passage ranking (2021) was another AI system Google applied to Search that year.
- The June 2021 core update rolled out in the same period.
The evidence
- Type
- ranking system update
- Started
- 18 May 2021
- Completed
- Not stated by Google
- Confirmed by
- Google announcement
Sources
- 1.MUM: A new AI milestone for understanding information - Google Keyword Blog, May 18, 2021Primary
- 2.How MUM improved Google Searches for vaccine information - Google Keyword Blog, June 29, 2021Primary
- 3.A guide to Google Search ranking systems - Google Search CentralPrimary
- 4.Creating helpful, reliable, people-first content - Google Search CentralPrimary
Frequently asked questions
Was MUM a Google ranking update?
No. Google announced MUM as a technology, an AI model it would apply to Search features over time. Its ranking systems guide says MUM is not currently used for general ranking in Search, only for specific applications.
When did Google launch MUM?
Google first announced MUM on 18 May 2021. It described its first application in Search, identifying COVID-19 vaccine names, on 29 June 2021. There was no single rollout date for rankings.
What is MUM used for in Google Search?
Google's ranking systems guide names specific applications, such as improving searches for COVID-19 vaccine information and improving the featured snippet callouts Google displays.
How is MUM different from BERT?
Google said MUM is 1,000 times more powerful than BERT, uses the T5 text-to-text framework, both understands and generates language, is trained across 75 languages and is multimodal, handling text and images.
About the author

Founder, UpgradIQ FZC LLC
Adam Hafez works on technical SEO and search measurement: how pages get crawled, indexed, ranked and now quoted by answer engines. He founded UpgradIQ, which reads Google Search Console and GA4 to tie ranking movement back to the changes that caused it. He publishes what the data supports and states the limits of it.
- Technical SEO
- Search Console and GA4 measurement
- Answer engine optimization
- Structured data
The briefing
Only what actually changed in search, delivered in full by RSS, Atom or JSON feed.
FollowRelated reading
AI shoppers now convert 54% better than non-AI traffic, Adobe finds
Adobe's May 2026 data shows AI-referred retail visits converting 54% better than non-AI traffic, a reversal from a year earlier.
Does adding JSON-LD schema raise AI citations? Ahrefs test
Ahrefs tracked 1,885 pages that added JSON-LD against matched controls. Here is the method, the citation changes by platform and the limits.
Report: Google AI Overviews now carry far more external links
Peec AI data says AI Overviews with external links rose from close to 0% to over 25%. Search Engine Roundtable shares it with doubt.
Google AI Mode adoption and zero-click rate, Semrush 2025
Semrush clickstream data on Google AI Mode's first two months: usage grew about 4x to just over 1% of sessions, and 92-94% ended without an external click.



