Contact us

Optimisation around user intent for voice search

news and insights
#Voice Search

With Google’s release of its specific search quality evaluations guidelines for voice technology, Artefact UK takes a look at the wider-ranging implications of the document for voice search.

Prior to December 2017, Google announced  that they’ve published a specific set of search quality evaluation guidelines for Google Assistant and Voice Search in order to give more insight into how its quality process works. The resulting PDF is rather short but extremely useful for anyone interested in this type of technology, from developers and academics to marketers seeking a better figurative footprint in the eyes-free search results.


The document references the existing Actions & Answers criteria for the Needs Met scale in the general search quality evaluation guidelines, so this isn’t particularly surprising. Google has always closely worked in line with user intent, but it’s particularly interesting to see them work on guidelines for voice action apps.


The examples for playing music rely on a human-level of common sense to determine how successful a response is. For example, when a user requests “play jazz” the app should play a playlist rather than a single song.

Even more conventional queries get this sort of treatment, for example when someone asks the name of the US president, they will want the current head of state. Or, if someone asks about the weather for the coming weekend, they’ll want to see a daily forecast for the whole weekend rather than just a general commentary.


“Play Mumford & Sons – Reminder” was a query that particularly caught the attention of our Artefact UK employees, as a testament to the sheer data parsing capabilities of the famous Google Knowledge Vault.

Not being huge Mumford & Sons fans, their initial thought around the intent was that someone wanted to set a song from the band as a reminder.

People with better (or worse) taste in music will, of course, understand that this query is asking for a specific track to play – so much for their human perspective on the query – while the example response given just opens a calendar app to set a reminder. The reason why they liked this example is twofold; firstly, it turns on its head the very old-fashioned idea of “latch onto the most obvious keyword”. Here “Reminder” is the name of a song, not the calendar-style activity. The second is that it is reliant on the kind of advanced semantic parsing capability, generally native to Google itself with its Knowledge Vault technology.

Our Monoprix case study shows that it’s not just music that voice search can get onboard with. Find out more about how we worked with Monoprix to produce a predictive shopping list.


Unless you’re a developer, probably not an awful lot right now – but quite a few brands are now starting to look seriously at the idea of integrating themselves into technologies such as Google Home and Alexa with custom skills. In our experts’ opinions, all brands should consider having it on their roadmap within the next twelve to eighteen months as the adoption rates of this tech aren’t showing any signs of slowing down. Brands that are too slow to embrace this growth stand the risk of being left behind.

Currently, there are some good implications in here that relate to preparing your brand’s content for Voice Search in general. There isn’t a single easy or technical way to “optimise” for Voice Search,but there are some general things to bear in mind.

First and foremost, rich answer results on Google surface commonly as responses to voice queries. While there doesn’t seem to be a direct equivalence (experiments at Artefact UK have shown that a voice answer doesn’t always match the rich snippet returned in “position zero” when sampling on a normal SERP) it is still worth keeping an eye on your footprint here. Make sure all your reference touchpoints like maps and Wikipedia are up to date and on-brand and see where else your content could be showing up for Q&A-type queries. Not everything gets a spotlight like Rodney McKay,) but that doesn’t mean you shouldn’t take steps to keep an eye on your SERP presence!

When it comes to trying to feature in more snippet, extensive experimentation and careful monitoring are strongly recommended for the best results. Content formatting and wording is a much stronger influence than technical mark-ups like So, create stuff that is fantastic for your users and focus on a “meeting intent” angle if you want to see progress. Always but always know what the purpose of your content is from a user’s perspective – are you answering a question (or questions), providing specific guidance on how to do something, explaining how something works…?

So, buzz your data, speak to your users (and customer service team), produce awesome stuff that meets the needs a real-world person might have in relation to your brand, and you’re already as well on the way as you can be to “optimising” for voice search in 2018.

Tools like are invaluable for generating ideas based on real searches, and of course, the outstanding rich snippet research done by the SERP sampling superstars at STAT. So go forth and develop your snippet presence, and don’t forget to have a good look at things like Actions On Google and Alexa skills development as well while you’re at it!


To discuss how voice search technology is evolving the way users and brands interact in the digital and real world, give us a call on +44 (0)207 607 5650 for a no-obligation coffee and chat about marketing and behaviour across all inbound channels.