Well, here we are. We’ve made it to 2020, the year that so many believed would be a milestone for voice search, because allegedly, 50% of all search queries by 2020 were going to be conducted via voice.

I think most people by now are aware that this stat is, to use the technical term, bunk. Many of them have been nice enough to link back to my own articles on the topic! (If you’re interested in reading about why the ‘50% by 2020’ “prediction” is so unfounded, you can do so here. The short version is that this statistic never existed as such; it also came from a prediction about voice search in China, which is a very different landscape to the west).

But today I want to go one step further and talk about why I think that talking about “voice search” at all in 2020 is missing the point.

I should preface this by noting that in this article, I’ll be talking about the voice and marketing landscape in the west, which has a different level of uptake, different players and different conditions to a country like China, where voice is much more well-established. For a look at voice in China, you can check out our recent piece on Baidu’s decision to exit the smart speaker price war.

Why using ‘voice search’ as a catch-all term is misleading

In my previous articles and reports on this topic, I’ve gone to lengths to point out that when many studies and articles talk about the rise of “voice search”, they might be talking about voice, but they’re not really talking about search. The below slide from Mary Meeker’s Internet Trends Report 2016 is my go-to example of this, because it so perfectly encapsulates the tendency to conflate voice commands, which are instructions given to a device using your voice to make it do things, and voice search, which is a web search carried out with voice as the input.

Figure 1: The rise of voice queries, from Mary Meeker’s 2016 Internet Trends Report

In this chart, Meeker uses three voice commands – ‘Navigate Home’, ‘Call Mom’, and ‘Call Dad’ – as evidence of a trend in voice search, even though clearly, none of them involves a search engine in any way. Meeker’s caption for the slide is a little more accurate in calling them “queries associated with voice-related commands”, but it still doesn’t get around the fact that these queries have nothing to do with search, and thus aren’t evidence of a trend in voice search.

‘Voice search’ is used as a catch-all term for voice interactions – it’s used variously to refer to voice search, voice commands, and voice apps like Alexa Skills and Google Assistant Actions; voice commerce is sometimes included, but is more often afforded a category of its own. So why is this a problem?

For one thing, it’s misleading. Talking about the ‘rise of voice search’ instead of ‘the rise of voice interactions’ mischaracterises the activity around voice devices as being search-based, when really, the majority of it is things like people using smart speakers to play music and set alarms. This is what enabled a transparently misleading statistic like the ‘50% by 2020’ stat to become so pervasive: if everyone is talking about this “voice search” thing, many commentators reasoned, surely there’s something to it?

Many studies and statistics (the Meeker Internet Trends Report above being just one of them) have added fuel to the fire by conflating different kinds of voice activity and labelling them all as ‘search’. These are then cited by commentators as evidence of the voice search ‘revolution’, and so on and so forth. Digging into the origins of these statistics would usually reveal that not all was as it seemed, but marketers are often guilty of citing a stat that supports their argument without interrogating it too closely – which is how it took everyone so long to notice that the ‘50% by 2020’ stat that is so popularly attributed to comScore did not, in fact, come from comScore.

Because voice is regarded as the domain of search marketing and SEO, the conversation around how marketers can take advantage of it has revolved around how to “optimise for voice” – even though for the majority of voice-related activities like playing music, asking about the weather, or setting a reminder, there’s no business opportunity and nothing to optimise for. (Creating a voice app like an Alexa Skill or Assistant Action is something businesses can do – but refer to my other article on why there are so few voice case studies for the reasons why that might also not be a great avenue to pursue).

We’ve reached the point where even Google’s Gary Illyes appears to be sick of the “voice optimisation” narrative, tweeting three days into January: “It’s a new year, maybe we can put the voice search optimization fad to sleep finally…”

Unsurprisingly, Illyes’ tweet prompted speculation from marketers as to his intended meaning. Barry Schwartz of Search Engine Roundtable pondered, “I don’t think he means that voice searches aren’t growing but rather that SEOs can’t optimize differently for voice search than what they normally do?”

But what if that is what Illyes meant – that there is little to no voice search opportunity to be optimising for? Would that come as such a surprise? All the evidence suggests that remarkably few people actually use voice to search the web – and companies like Google know it. In 2016, Sundar Pichai announced at Google I/O to great fanfare that voice made up 20% of all search queries in the Google app and on Android in the US. That was coming up on four years ago, and Google hasn’t made any moves to update the statistic since.

In a previous article, I estimated that if one in every five mobile searches really is a voice search, this would only add up to around 12% of searches, or 420 million queries – and that’s if we ignore the other constraints on Pichai’s statistic (“in the Google app and on Android” and “in the US”) and are as generous as possible about what constitutes a ‘query’, knowing that many of these aren’t queries that can really be ranked for. In short, it’s not a lot of searches, and we’ve had no indication that the percentage has grown since 2016.

What about smart speakers, those great agents of the voice-first revolution? According to the most recent figures from Voicebot’s Smart Speaker Consumer Adoption Report 2019, 84% of smart speaker owners have tried using their device to ask a question at some point, but only 37% use this functionality daily. The percentage is even lower for product info searches (the kind most potentially valuable to brands, since again, not all queries can be ranked for), which only 41% of smart speaker owners have ever conducted, and just 11% carry out daily.

The study also found that overall smart speaker usage is declining, with daily smart speaker users declining by 16.2 percentage points between 2018 and 2019 and the percentage of smart speaker owners who use their device “never or rarely” increasing by 13.6 points – which translates to a much smaller pool of potential searchers.

Voice in 2020 is (still) not about search

You can’t blame users for passing up voice search when its user experience is not exactly stellar – on mobile the only real difference is the input method, with the output being a regular screen of mobile search results, perhaps with the top one read out. On a smart speaker, voice search is still largely a single-result experience, with no meaningful onward journey; Google’s ‘continued conversation’ feature introduced last year makes it possible to ask follow-up questions, but there’s still no such thing as browsing the web via voice.

Google has made one or two passes at features that might have improved the experience of searching the web via voice, if they were implemented in the right way – but it hasn’t followed through. In July 2018 Google announced Speakable Schema, a markup that web developers can add to a page to indicate the sections that are most suitable for being read aloud by a voice assistant, and trialled it with a number of news publications in the United States. Last December, Danny Sullivan confirmed on Twitter that Speakable is no longer restricted to news content – but emphasised that it remains in beta, and that “using Speakable markup on any site isn’t a guarantee that the Google Assistant will always use it”.

The lack of an official announcement on this from Google is noteworthy: if Google wanted to showcase its efforts to build a more robust voice search experience, it would have made more of the fact that Speakable is now available for content of all kinds – and probably would have put resources into bringing it out of beta.

At CES 2020, Google also previewed a feature called ‘Read It with the Google Assistant’ (set to launch later this year) that will enable the Google Assistant to read articles and blog posts aloud to the user, ‘audiobook’ style. This could potentially have been a huge development for voice search and the ‘audio web’ – but there’s no connection to search. Instead, the feature only works if the user first pulls up the article on their phone, before asking Google to read it aloud.

It’s tempting to paint these as positive developments for the eventual future of voice search, but the fact that Google has been quiet on Speakable Schema and not drawn any link between voice search and Read It with the Google Assistant speaks volumes (if you’ll pardon the pun). A cursory glance at the full list of announcements from Google at CES shows where Google’s focus really is: building on smart home functionality and integrating the Google Assistant with devices, not the web.

As it stands, there’s little desire among consumers to use voice search and while it’s possible this could change if it provided a better experience, the work this would involve on the part of Google (or, hypothetically, Amazon or Apple – but neither company is particularly known for search) vastly outweighs the potential benefits. Voice search was an exciting novelty in 2014 and even 2016, but in 2020, it has receded far into the rearview mirror.

Google isn’t talking about voice search. Consumers aren’t talking about voice search. Most of the time, when marketers talk about “voice search”, they aren’t even referring to voice search.

So maybe it’s time to retire the term for good.