Google MUM: What the Multitask Unified Model Means for Search

Google uses its annual I/O developer conference to show off its latest innovations, and at I/O 2021, in May, one of them stood out as something that could change the way we search for good. Meet Google MUM, a new AI technology that Google says will transform how people find information on the world’s most popular search engine. Here’s what we know so far.

MUM: Built to Answer Complex Questions Better

After canceling I/O in 2020 because of the pandemic, Google came back in force at I/O 2021 with a technology developed in Mountain View. Coincidence or not, it unveiled MUM on May 18, just nine days after Mother’s Day.

MUM stands for Multitask Unified Model. Its goal: to understand and connect several types of information at once, whether text, images, and eventually video and audio, so it can answer queries made to Google more precisely. And it’s trained across 75 languages.

With MUM, Google is going after complex, long-tail questions that are usually too specific for a single search. The aim is to give people a simple answer when the question isn’t simple. Google based the idea on a clear finding: people need, on average, eight searches to complete a complex task. For Google, that’s far too many.

Because it’s built on artificial intelligence, MUM is likely to change the way we search for information and, as a result, the way businesses present it. MUM is designed to help searchers, but it will also affect the businesses on the other side of the screen.

Google’s Own Example

To make its case, Google shared this sample query with its virtual audience:

I’ve hiked Mt. Adams and now want to hike Mt. Fuji next fall. What should I do differently to prepare?

Today, a question like this takes several separate searches: comparing the elevation and difficulty of each climb, checking average temperatures, figuring out what gear to bring, and more.

If you asked a hiking expert in person, you could ask one simple question and get a complete answer. That’s the gap MUM aims to fill, but on the web. It’s meant to understand exactly what you’re asking, including everything implied behind the question, and then answer it more fully.

1,000 Times More Powerful Than BERT, but Still in the Works

Google isn’t just promising better answers. According to Prabhakar Raghavan, Google’s senior vice president in charge of Search, MUM is 1,000 times more powerful than BERT, the language model Google introduced in Search in 2019. It’s a bold promise, and we can’t wait to see it in action.

When Google announced MUM in May 2021, however, it gave no launch date. Google described a period of internal testing and refinement, since the technology is still in its early stages.

What’s clear is that MUM is another reminder of how much Google shapes the way we all find information online.

Where Things Stand in 2026

Google has since put MUM to work in specific features rather than in general ranking. Its guide to Google Search ranking systems lists MUM but notes that it isn’t currently used for general ranking in Search, only for specific applications, such as improving featured snippet callouts.

Get Your Website Ready for What’s Next in Search

As you can see, search is constantly evolving. The current algorithm has barely settled in before the next one is ready to take its place. That’s why it pays to keep up, because online performance has become one of the most effective ways to stand out from the competition. Our guide to SEO in the AI era covers where search has gone since.

At Référenceur, we enjoy putting every new Google development to work to give our clients more visibility, and ongoing training is part of everyone’s job on our team. We’re a French-speaking agency from Belgium and Luxembourg, and we help U.S. businesses remotely with SEO services and paid search campaigns.

Want to improve your visibility and get more out of Google’s tools? Get a free assessment.