Is the hype surrounding voice search deserved? There’s no shortage of advocates on voice search, but I hope that we can help cut through some of the noise with this SEO-focused guide to voice search optimisation. In this post, we’ll explore the statistics, user behaviour, and ways in which you can prepare your site for voice search in 2019.
Current state of voice search
It’s been a few years since Google released statistics on the share of voice-activated searches, however back in 2016 Google revealed that 20% of mobile searches were made via voice.
Since then, Google’s been relatively quiet on releasing any data around voice search adoption. However, they shared in 2018 that nearly three quarters of US owners of smart speakers use the device as part of their daily routine.
In order to dissect voice search usage, we need to understand what’s driving each search result. Here’s the breakdown of the top types of voice search and the search engine responsible for delivering the search results:
- Google Assistant = Google Search Results
- Apple Siri = Google Search Results
- Amazon Alexa = Bing Search Results
It’s important to note that Siri uses Apple Maps to deliver localised search results, which Google Assistant uses Google Maps to deliver those local results. However, both use Google to deliver general search results. This may change as their relationships and the technology evolves, but this is the current state in 2019. Meanwhile, Amazon uses Bing to deliver all search results as it is currently entrenched in a very competitive battle with Google on many fronts.
Regardless of engine, they are all focused on delivering the best result for their users, and rightfully so. Google’s head of search in the UK, Alessandra Alari, recently stated:
“For businesses to get the most out of voice search they need to strive to be as assistive as possible. This means considering not only what questions consumers will be asking, but also how, when and where they are asking them.”
Let’s make a mental note of Mr Alari’s guidance highlighting the need to consider “questions consumers will be asking,” as we’ll come back to questions later in the guide. But let’s back up and explore how we got here.
Very brief history of voice search
In 2013, Google’s Hummingbird algorithmic update allowed the search engine to shift focus from keywords to natural language processing, which allowed Google better understand the search intent of the user.
This update was later followed by Google’s RankBrain in 2015, a machine learning algorithmic update focused on interpreting search queries to provide the most accurate result. While Google’s Knowledge Graph began collecting data back in 2012, these two algorithmic updates allowed for Google to better understand all types of search queries, serve more accurate results, and opened the floodgate for Google’s Knowledge Graph to populate search results with “rich results.”
Officially dubbed “rich results” by Google, these include answer boxes at the top of search results, accordion-style Q&A “people ask ask,” the Knowledge Graph panels for search entities (i.e. person, place, or thing), and many more enriched search results.
In the voice-activated search result below, Google Assistant reads out the first line from Iggy Pop’s Wikipedia entry.
According to Wikipedia, James Newell Osterberg Jr., better known as Iggy Pop, is an American singer, songwriter, musician, record producer, and actor.
Then If I ask, “how old is he?” Google now understands that he is Iggy Pop as a result of the previous search and can tell me he’s 72. Notice the lack of attribution below.
Why does this matter? As we’ll explore further in a bit, this led to Google’s ability to serve bite-size nuggets of information in order to answer your question directly in search results, without you having to click a result. This fed perfectly into serving content on voice-assisted search devices by removing the need to click and only requiring the user’s voice.
This update also allowed for their algorithm to more accurately answer queries which are more conversational in tone, a highly common user behaviour with voice-activated search we’ll explore next.
User behaviour with voice search
Let’s face it, typing on a phone is a pretty horrible experience. This, along with the advancement of technology supporting it, is certainly why voice search is on the rise.
User behaviours on voice search carry these five key attributes, as summarised by Heather Physioc:
When users need to type their search query, they tend to do so in a much shorter-form. This might be because we’re lazy when it comes to typing, or because we’ve been conditioned for decades to search for keywords.
With voice search, users have the tendency to use natural language in searches. We are much more likely to ask a question or use a conversational tone when speaking to voice-assisted search devices.
To point out the obvious, there’s two different ways in which a voice-activated search is conducted by users:
- Users search on their mobile phones using their voice
- Users search on their smart speaker using their voice
Both of these methods of voice search have similar initial user experience interaction but different search results formats. Users of mobile voice search are presented with visual results on their phone along with an audio answer reply. Users of smart speakers (without screens) are presented with an audio answer reply only. This means they only have one option with no other search results to browse.
Earning the result the search engine reads aloud is paramount to visibility on both device types. Google search results habitually use rich results for the the audio reply, which presents an opportunity to “appear” prominently in search results.
Let’s dig deeper into each method and the implications they have for your site and how to optimise for them.
Voice search results on mobile devices
Rich results are the most common visual and audio result in voice search results on mobile devices. Also known as “position zero,” this result appears above all other organic search results and isn’t necessarily the first search result. Oftentimes, a rich result can be the second, third, or even fifth organic search result. This leaping of search results isn’t limited to organic results, when we consider voice search result.
For example, let’s say your page ranking in position three manages to earn a rich result at the top of organic results. Let’s also say the search result has two pay-per-click ad results at the top of search results, the rich result at “position zero” is still below the ads in the search results results. However on voice search results, earning the audio reply will placing it in in first place, leaping over both paid and organic results.
Rich results are earned through a combination of content relevance and site authority, but more importantly, the search engine’s ability to find, process, and extract the information off the page. Due to the authority and broad relevance of Wikipedia site pages, there’s no surprise that they dominate rich results for general knowledge queries.
However, longer-tail and more specific searches are commonly answered by all sorts of sites, from publishers to user-generated content.
Voice search results on smart speakers
With no visual result available on the most common types of smart speakers, the search engine returns only one result, commonly the rich result. Earning the coveted “position zero” is even more critical on these screen-less devices as there’s no visibility for any other search results. However, users of these devices are rarely doing anything more than gathering information in most search results.
Despite massive growth and adoption in the category, usage for transactional queries is very limited. This was illustrated by recent UK-based study by eMarketer:
“Despite the growing capabilities of smart speakers, people still tend to use them for basic functions: listening to music or podcasts, listening to the news, asking for weather and traffic, and asking general questions.”
The likelihood of queries resulting in transactions, outside of Alexa-based and Amazon-isolated situations, is few-and-far between. Rather than purchasing on their device, users will add-to-cart on their smart speaker and complete the transaction on a traditional screened device, according to eMarketer. However, with the addition of screens to smart speakers, this is likely to change in the near future.
How do you earn a rich result?
Rich results remain the best method to own voice queries, both on screened and screen-less devices. In order to earn a rich result, there’s key elements in the formula to employ, but also a rather significant a dash of luck.
Keyword research for voice search
Since the majority of voice searches are natural language queries, questions searched by users are a good starting point. These are often overlooked by competitors and can a be great way to answer your target audience’s questions exactly as they are asking them.
Step 1: Use a Google suggestion tool to find questions around a topic.
Google suggestion tools take advantage of the autocomplete feature in Google search results. This is where when you’re typing in a search, Google makes suggestions on what to search based on popular or relevant searches on the topic you’ve entered thus far.
These Google suggestion tools utilise Google’s Autocomplete API to pull in all of these suggestions around a topic you enter into the tool. My personal favourite is KeywordTool.io, specifically for its sheer speed and no-nonsense approach. There are a handful of tools that that utilise the same data and present it in different ways, notably AnswerThePublic.com. We’ll focus our example above using KeywordTool.io.
Step 2: Enter your search term or topic on KeywordTool.io. Notice all the other suggestion tools you can use beyond Google.
Step 3: Click the “Questions” tab and copy all results to your clipboard.
Step 4: You’ll notice that the search volume for all the questions are blurred out. We don’t have a paid version of this tool, but with it you can pull in Google’s search volume directly into their results. In order to manually get Google’s search volume, take the copied keywords and enter them into Google Ads Keyword Planner.
Step 5: Select “Get search volume and results,” paste in your copied questions, and hit “Get Started.”
Step 6: Select “Historical Metrics” and sort by search volume. Note: Google Ads now requires an active account to display finite numbers rather than ranges.
Step 7: Looking at the top searched terms, manually search all of the relevant terms to you in order to identify which search terms do not have rich results. You may be able to hijack a rich result from a competitor, however it’s often easier to tackle those search which don’t already have a rich result.
In my research, I found the question “what dslr should I buy” had no rich snippet and earned 210 searches per month:
No rich result appears at the top of the search result, which represents an opportunity to target this keyword directly with an article or a dedicated page.
So we’ve covered how to identify keyword targets for rich results using natural language queries, which match user behaviour on voice search. Let’s shift to your web property and illustrate way you can earn a rich result, and more generally, how we can optimise your content for voice search.
Optimising your site’s pages for voice search
Let’s start with the question we defined above. In order to rank a page for this keyword, we will need to create either a dedicated landing page directly targeting that question or a page very closely aligned with a slightly broader topic.
Some attempt to capture rich results by adding questions to FAQ pages. While this might work in some cases, FAQ pages are often too broad to be relevant to such a specific query. Remember that rich results are also an actual search result, meaning that if it would be unusual for the page to appear with the other results, then it’s certainly not likely to rank amongst for either a rich result or these specific organic results. While FAQ pages do fulfil a purpose, we more commonly recommend our clients to create dedicated landing pages or thematic pages to target these queries.
Once you’re ready to write your content, utilise a conversational tone rather and simple paragraph structures. A comprehensive search results analysis by Backlinko discovered that the mean reading level of content that won rich results was at a 9th grade reading level in the US, the UK equivalent of Year 10.
When it comes to answering questions directly, Backlinko data suggests the answer to be concise at an average of 29 words in length. This means that if you have a a paragraph or section of your page dedicated to directly asking the question (ideally wrapped in an appropriate heading tag) and then a section of body copy answering it, you’re to keep it relatively concise. This seems logical as Google can only display so many characters in a rich result. Using an incomplete string of text that is truncated may not fully answer the user’s query and be a poor search result experience.
According to the Backlinko study, page speed plays a major factor in awarding rich results. This comes as no surprise since Google switched to a mobile-first index and page speed became a ranking factor in July 2018. The page speed mean for rich results was 4.6 seconds, which is 52% faster than the average page.
Other top tips recorded in this Backlinko study were:
- Ensure your site utilises HTTPS as 70% of Google Home results did. If you’re not already, please do as this has been a ranking factor since 2014.
- Google appears to favour longer-form content for sourcing rich results, with the average word count hovering around 2,300 words.
- You can feel comfortable using desktop results to guide your research as roughly 75% of rich snippets matched mobile results, but recommend double-checking any decisions on mobile.
Utilise Structured Data
Additionally, you can markup your page with Structured Data, and more specifically Schema, an initiative launched by the search engines themselves including Google, Yahoo, Microsoft, and the Russian search engine Yandex.
Structured data is a markup that pairs information with data values to help search engines categorise and better understand your content. Browse Google’s gallery of structured data available for all types uses, including products, recipes, events, articles, and organisational Schema.
In response to the rise in voice search, Google and partners have launched Speakable Schema, which ” identifies sections within an article or webpage that are best suited for audio playback using text-to-speech (TTS).” This markup will enable Smart Speakers to read more content aloud, such as new articles. Currently, this Schema markup is still in BETA and is only available to Google News Producers. While still in its infancy, Speakable Schema will very likely be a key component in the near future and the landscape continues to evolve.
Local voice search
While this voice search guide is certainly not exhaustive, it would not be complete without discussing local implications.
With a dramatic rise of “near me” searches, partially enticed by Google suggestions, ensuring your local presence in search is essential if you have physical locations open to the public or offer local services.
In order to ensure you’ll show up in these localised searches, it’s critical to claim and optimise your local business listing in GoogleMyBusiness. This includes ensuring all your details are correct and that you’re utilising the optimal business category and appropriate photos. There are a host of activities to further optimise and promote your GoogleMyBusiness listing such as posting on your local listing and engaging with users to ensure an optimal reputation is managed.
Beyond GoogleMyBusiness, we highly recommend local-focused business utilise a local SEO tool to manage listings across a host of directories and other map-based sites. Working with the big four data aggregators, these local SEO software solutions ensure your data is correct across all these directories without having to manage them directly. It’s a massive time-saver, especially if you have many locations.
Depending on your needs, you might be more interested in BrightLocal for only a few locations or if you would like to manually manage some of the work. For the enterprise route, we recommend using a global solution such as Yext or Uberall. Regardless of which software you select, you must go through one of these software solutions in order to work with the data aggregators.
With voice search search continually on the rise, it’s more critical then ever to be paying close attention to how users are searching on mobile devices and smart speakers via voice. The hype is real, but less so for commerce.. for now. Hopefully you found this guide helpful in identifying the tactics you need to prepare your site for voice search in 2019. To Learn more about our SEO services, get in touch.