What Google’s MUM update means for international SEO
Google announced its latest update – MUM – in May 2021. In September, Google released previews of how the update could affect how people search for information. It is clear that MUM’s scope is huge and will potentially impact not just every search result but how we use Google, both as consumers and marketers. Read on to find out what the update means for multilingual websites and international SEO.
What is MUM?
MUM stands for “Multitask Unified Model”. It’s an AI model designed to use machine learning to revolutionise how Google understands search queries and online content – and how it delivers search results to users.
MUM isn’t one single technology. Instead, it’s AI used to perform different tasks (hence the name, ‘multitask’) in what promises to be a huge leap forward in search.
What will MUM actually do?
If you asked Google to pinpoint the most important aspect of MUM, they would probably say it is the natural language processing (NPL) element. By understanding language in a more semantic way, similar to a human, Google is able to understand the context of a search much better. BERT – Google’s previous language processing technology – was already making strides with this, but Google claims MUM is a thousand times more powerful.
With MUM, Google will understand what people want to know better than ever, as well as understanding the information on the web in greater depth too. This will make it possible to find relevant information that may not mention a user’s search term or even synonymous terms but does answer their query.
At the moment, if you have a nuanced or complex question, you probably have to carry out multiple Google searches before you get a complete and well-rounded answer. (Google states that on average, more complex tasks involve eight search queries.) MUM aims to reduce the number of search queries by giving users more nuanced, sophisticated answers from the start. It will do this in four ways:
- Being language agnostic – i.e. searching for answers from 75 different languages – because the best answer may not necessarily be in the searcher’s own language.
- Understanding information across format types – looking beyond the written word to place greater emphasis on image, video and audio content as well.
- Providing predictive answers – i.e. anticipating the full, multi-layered search intent behind a user’s query so it can provide tailored answers and related helpful content without the user needing to ask a series of subsequent questions.
- Combining multiple sources into a unique, tailored answer – producing unique prose that takes information from multiple sources and summarises the most important and most relevant points.
On the language point, to be clear: Google’s ability to understand the semantics behind language means that it will be able to combine it with other sources to return a result in the user’s own language.
It isn’t only Google which thinks NPL is the future. Baidu’s ERNIE 3.0 NPL engine is trying to achieve the same objective and by understanding English as well as Chinese, will vastly increase the amount of information available to users.
These generated answers probably won’t replace traditional results but are likely to replace or evolve featured snippets. While details are yet to be announced, this is the area where Google has provided the most detailed preview and may be one of the first changes we see in SERPs when these ‘super snippets’ roll out.
What do MUM super snippets mean for international websites?
Currently, Google offers users featured snippets at the top of its results page. These are brief excerpts from a website which Google automatically pulls to provide quick answers to a user’s question. Featured snippets aim to keep users on the results page for longer (rather than clicking through to other websites – hence the term, zero click searches).
MUM will expand and evolve featured snippets so they appear richer with more detailed, tailored information. We don’t yet know exactly what these enhanced snippets will look like, although we do know that MUM will be able to create unique tailored content augmented with different media types which will vary depending on the relevance to the search term. The predictive answering part of MUM will be able to deal with complex queries by giving extra information that a user might need and providing options to explore further. Crucially, sources for the content will be included so there is still some opportunity to gain visibility.
This sounds like a lot of information to fit into the SERPs, which is why we expect this to be organised into different tabs, similar to some featured snippets Google experimented with for Covid-related queries. Even so, we can expect these new snippets – especially on mobile – to push other search results further down the page.
Super snippets will also see MUM using sources written in other languages, translated into the user’s language. This will open up more local and insightful information than any previous Google update.
For lesser-used languages that have fewer results compared to a major language such as English or Spanish, the total available knowledge is set to increase drastically. This may also help Google’s penetration into markets with strong local search engines such as Naver in South Korea and Seznam in Czechia.
For other languages, such as English, having access to local knowledge for travel, news, sport, food and recipes from reliable, expert sources could greatly improve the quality of these results.
Let’s look at an example:
At the moment, if you search for “How much does it cost to enter the Crystal Palace in Madrid” in English, you receive basic information as the top answer – namely, that it’s free to enter. However, this is from a forum comment on TripAdvisor, so may not necessarily be trustworthy.
However, the same search query in Spanish – “¿Cuánto cuesta entrar al Crystal Palace de Madrid?” – yields much richer results from a range of expert, local sources – for example, how you can get there on public transport, what else is going on nearby, what are the opening hours, user reviews, and so on. MUM will be able to present this information to English users too – by translating into English – and presenting it in a neater, more tabulated way than the current SERP.
However, Google recognises there will be user cases where this is less useful. For example, if you search for “supermarket reds” in English, you’re presented with a list of good red wines which you can buy from local supermarkets. Google recognises that search results from other languages – such as Spanish – “mejor vino tinto de supermercado” – showing what wines are available in Spanish supermarkets – is not useful to users based in the UK, and so won’t show these results.
How will MUM affect international businesses?
For international marketers, MUM has big implications. Two in particular stand out:
#1: More localised insights:
Breaking down language barriers will allow users, SEOs and businesses to see more localised insights and responses. For example, local people create online reviews – across categories such as tourism, e-commerce, B2B, financial services and many more – yet at the moment, users may miss what could be the best answer to their review-style questions because it’s in another language.
#2: More competition to be a source:
With MUM pulling information from multiple languages, your existing featured snippets may no longer be a source for MUM’s super snippets. A better article could exist in another language. It seems likely that Google will also use MUM’s NLP to start pulling in more results in other languages to the traditional results and then translating pages on the fly, if they are a better result. More localised insights could displace multinational sites, such as TripAdvisor in the previous example, and see them lose visibility to local competitors.
The most impacted searches are likely to be in less-spoken languages which have fewer sources, where the best answer is probably in a major language. Let’s look at some hypothetical websites to assess potential opportunities and threats:
I operate a multilingual website:
- You’re winning if your content is already fully localised (not just translated), since post-MUM you will still retain an advantage. This is because your content will reflect linguistic and cultural nuances across markets in a way that machine-translated content can’t.
- You’re losing if your multilingual content is based entirely on auto-translated or direct-translated content, because any multilingual advantage you might have had previously is now gone – since Google will now show auto-translated content from other sources if they offer a better answer to you.
I operate a monolingual website in a major language (e.g. English or Spanish):
- You’re winning if your site strength means you have more opportunities to display in search results for languages with few good sources.
- You’re losing if you now face more competition in other languages, and local expertise in other countries could mean you lose geo-relevant search visibility.
I operate a monolingual website in a less-spoken language (e.g. Vietnamese or Greek):
- You’re winning as you could now become a source for any search – mainly geo-relevant searches – broadening your international visibility.
- You’re losing if you now face more competition and could lose out to stronger sites. For example, imagine a Vietnamese website which has content about theatre shows in London. At the moment, it doesn’t face much Vietnamese competition but post-MUM, will be competing with numerous English-language sites which focus on London theatre shows – losing traffic in the process.
When will MUM be rolled out?
The algorithm is currently in testing and the precise roll out is TBC. Google have recently cited 2022, though whether this will involve some features or a full roll out isn’t clear.
Google knows that with machine learning, it is vitally important that only accurate, reliable information is used as inputs, to ensure accurate information as outputs. So with such a complex system, they are sure to test it rigorously before rolling it out.
Google’s previous algorithm – BERT – was in testing for about a year, with roll out taking another year thereafter. And that was just for English: with MUM encompassing 75 languages, we can expect it to take longer.
Google has promised other MUM-powered features and improvements soon. In the meantime, Google will continue to test and refine MUM to address various concerns, including applying its latest research on how to reduce the carbon footprint of expansive machine learning training systems.
What should international SEOs do now to prepare?
Ultimately, the steps needed to prepare for MUM are what SEOs should be doing anyway – namely, creating high quality content which answers audience intent and is presented in a way that is easily found by search engines:
High quality content:
- Content must answer your audience’s questions – so it’s essential you understand what your audience searches for and how they search for it. Bear in mind this may vary across cultures – use a Local In-Market Expert to guide you.
- Written content – e.g. blog posts and articles – is still important, but given MUM’s greater emphasis on image, video and audio content, think about what multimedia formats you can create to help your audience.
Localisation is still important:
- ‘Localise don’t translate’ has always been best practice and will be your key differentiator post-MUM. Understanding how different cultures conceptualise your product or service, how they search for it and how they buy it – this is still the key to unlocking the biggest international opportunities.
- Investing in good local copywriters – i.e. native speakers based in-market – will continue to pay both SEO and brand dividends.
Mark up your content properly:
- Remember to add structured data to your page to give clues about the content.
- Pay attention to context and semantic relevancy to capitalise on MUM’s emphasis on comparisons and related content.
- Use SEO principles to format your content – e.g. using headings, bullet points, considering the optimal word count, linking to good quality sources, and so on.
Whilst MUM aims to remove language barriers, multilingual SEO still needs to be part of your strategy. MUM will increase the regional power of content, so brands which understand how best to leverage it will ultimately be the winners.
Google’s MUM update: Summary of key points
- MUM stands for “Multitask Unified Model”.
- It’s an algorithm which aims to search the internet in 75 different languages including text, image, video and audio content.
- MUM aims to reduce the number of search queries by giving users more nuanced, sophisticated answers from the start.
- It will do this by being language agnostic, understanding different media types, and providing tailored, predictive answers to users.
- Featured snippets will evolve into super snippets – i.e. snippets at the top of the SERP which contain richer information from numerous sources, organised into tabs.
- These snippets will take up more of the page, pushing other results further down the page, especially on mobile.
- We can expect more zero click searches as a result.
- The update will mean more opportunities for localised insights (for users and businesses alike) as well as more opportunity for brand discoverability.
- The most impacted searches will be in languages which have fewer sources – since it’s likely that the best source for various topics will be in a major language.
- Localised content will continue to have the edge over machine-translated content – since it captures linguistic and cultural nuances in a way auto-translations can’t.
- Google’s focus on the importance of E-A-T – expert, authoritative and trustworthy content – is more important than ever.
. . .
Want to prepare for MUM now to ensure you stay ahead? Oban can help – get in touch to find out how.
Oban International is the digital marketing agency specialising in international expansion. Our LIME (Local In-Market Expert) Network provides up to date cultural input and insights from over 80 markets around the world, helping clients realise the best marketing opportunities and avoid the costliest mistakes.