SEO and Voice Search Goes Together Like Sardines and Peanut Butter

Author Image
by Dan

Voice search is all the rage. To hear it from a marketing or search technologist, the voice search revolution is on us. Lowering barriers to access and radically changing the way we interact with devices. There is only one problem, take it away Dr. Malcolm

I hate to break it to everyone, but voice search is not good. Like straight up not good in terms of being able to consistently provide answers to search queries in ways that are sensical and replicable. I get it, it’s super fascinating and future tech to be able to interact with a machine via your voice. All us nerds want to live in the world of the Star Trek computer, at least me and my boy Amit Singhal do, but this is the .001 version of that concept and we all need to embrace that.  At the end of 2016, we provided everyone at Local SEO Guide Google Home’s to do some voice search research, something that Andrew and I have been playing around with a bit on our own for a couple of years now. These are their stories.

Results Hodgepodge

Voice search regularly provides a non-intuitive user experience or fails to deliver relevant results. Now I can just hear people saying:

What are you talking about Dan? How can talking not be an intuitive user experience?

Pretty simple answer random internet stranger, voice search engines still have a hard time with natural language processing. There, I said it. Here is an example for non-standard words e.g. a Japanese restaurant called “Tanuki”:

  • Thought searcher was saying “The New Key” the first several times I requested it.
    • Even tried “The New Key Tanuki” and it recognized both as “the new key”
  • Eventually found it using a directions query asking “where is the restaurant tanuki?”
    • Even though searcher is in Portland it gave them directions to a restaurant in Northern California
  • Had to clarify “tanuki portland” for those directions

Even if it gets things right 90% of the time, which I doubt it does but would need to do research at scale, that’s a 10% fail rate. If your web interactions failed 10% of the time, how annoying would it be to use?

Also, it just isn’t that great at returning relevant search results. Just looks at this string of fails that Siri had to some pretty basic natural language questions:

“Where can I buy a paddle board?”
Voice Search Query For Paddleboard

“Where can I buy a standup paddle board?”
Standup Paddle Board Voice Search Query

And my favorite, asking how to make Chicken Curry provides the Wikipedia page, not a recipe. GG Siri!

Chicken Curry Voice Search Query

It also often returns less valuable results than the web (desktop or mobile) searches, though that is something I will dive into at a later date.

Getting Better Results Means Searching Differently

One of the things we have noticed in our research is that adding more semantically relevant descriptors around commands and voice search queries provides better results. This is great for a power user, but bad for casual users who want to use NATURAL LANGUAGE, not be trained in the search protocols and interpretive quirks of voice search. Here is a pretty common example in regards to this:

Ex: Tried to listen to Chris Thile

    • Results from “Play Chris Thile”:
      • Interpretation: Play Chris Bailey
        • Response: Alright, check out this Chris Bailey radio mix on Google Play Music.
      • Interpretation: Play Chris Farley
        • Response: My apologies… I don’t understand.
    • ` Results from “Play Chris Thile Bluegrass”
      • Interpretation: Play Chris Thile Bluegrass
        • Response: Since I couldn’t find the album Edgar Meyer & Chris Thile in your music library, here’s a similar mix on Google Play Music.

So here is a question, why did adding the music genre mean that Google should look for it in my library? And why is it doing that since I have Spotify set to my primary music application? The world may never know. Also, this is the type of user experience that would royally piss of someone like my mom (and probably your mom too). If voice search is aimed at the lay user versus the power user, expect them to regularly have a horrible time with your brand during a good percentage of voice search interactions.

Sorry to throw a bucket of water on the technologists and marketing technologists, but voice search just isn’t ready for the primetime as a marketing channel/strategy just quite yet. We are going to dive deeper into this subject in the coming weeks. Up next, voice search results don’t different from web results… Or do they?


  • Twitter
  • Facebook
  • Mail
  • LinkedIn

this content