Ever wonder if your device is actually listening to everything you say? In the world of digital advertising, the biggest brands and agencies are constantly clamoring to continue to target in more precise and innovative ways. Given that, could listening into our conversations be the next big thing? 

Brands such as Apple, Amazon, and Google say no. Although perfectly legal for platforms to harvest this data (a.k.a. listen in on your conversations), they attest it is purely for enhancing algorithms and functionality, and not for marketing purposes.

But the fact is, a Bloomberg study reported that Amazon dedicates teams of employees to listen and record data that is captured in people’s homes. Understandably, with the rise of smart virtual assistants such as Siri and Google Home, users have become more and more concerned about their privacy.

To address these concerns, tech giants have publicly stated that these devices will only be listening when specifically called on by their activation words, such as the oh-so-familiar ‘Hello Alexa’.

In a memorable event, Facebook CEO Mark Zuckerberg, has even gone so far to state in front of congress that these concerns are a “conspiracy theory” after direct questioning of device listening.  But somehow, we needed a bit more convincing. So a group of University of Texas at Austin Graduate students decided to take it upon themselves to run through a series of listening experiments.

How We Ran the SmartSpeaker Qualitative Research Study

Nine graduate students tested device listening with 3 different key topics; babies, guns, and retirement. These key topics were chosen based on their uniqueness to the group — none of the testers had searched or normally spoke about these topics.

The graduate students were divided into three groups, each testing for a specific count of topic mentions (10 times, 15 times, and 20 times), that would be said out loud in the vicinity of their digital assistant, without using the activation words.

In other words, these topics were discussed without waking up the SmartSpeakers by saying “Hey, Google,” “Hey, Alexa,” or “Hey, Siri.”

Participants used sample scripts like “I wonder what the best stroller for my baby is,” or “how much do retirement homes cost?” Devices used for testing were Siri, Google Home, and Alexa.

In addition to the topic study, the students also tested for listening devices being able to pick up a foreign language. Similarly, the language chosen was not one any of the students usually spoke or searched for — Russian. The group wanted to understand if these would affect any advertising being displayed in foreign languages, or being advertised language-learning tools.

We expected to see advertisements related to our topics of conversation after speaking around non-activated SmartSpeakers and our phones with Siri enabled. For example, we thought we’d see ads for strollers and diapers for our baby topic, retirement homes and savings for retirement, and so on. 

What Were the Results of the Google, Siri, and Alexa SmartSpeaker Research?

So, what did we find? Generally, our SmartSpeakers may not be picking up as much information as one might think, but we’ll get more into that later. Our findings topic-to-topic are as follows: 

Guns:

  • 2 out of 9 participants received ads for gun-related topics during the testing period.
  • 2 ads served overall.
  • Ads were served to 1 Google Home user, and 1 Siri user.

We knew advertising for guns is a no-no on social media platforms, but we expected to get related topics. One of the two ads received was a gun-safety ad promoting gun safes in homes with small children. The only ad our male participant received was gun-safety related. 

Retirement:

  • 2 out of 9 participants received ads for retirement during the testing period.
  • 4 ads were served overall.
  • Ads were served to 1 Google Home user, and 1 Siri user.

While the key phrases we used were about retirement homes, the ads that were served covered topics like investing and financial planning. 

Foriegn Language:

  • 2 out of 9 participants received ads related to foreign languages during the testing period.
  • 4 ads were served overall.
  • Ads were served to 1 Google Home user, and 1 Siri user.

While none of our participants received any ads in Russian, they did receive ads for foreign language learning services, like Babbel, and Lingopie.

Babies:

  • 4 out of 9 participants received ads related to babies. 
  • 28 ads were served overall
  • 3 out of 4 users were using Siri, the fourth was using Google Home. 

The 28 baby-related ads were covering a wide variety of topics, like toys, baby monitors, pregnancy, and even birth-control. While some of the ads weren’t specifically about babies, they included language like “baby” or “kids.” Every participant that received ads for baby-related topics were female, and 23 of the 28 baby ads were served to just one of our participants. 

Overall, participants reported 2-4 ads total for three of our four topics (guns, retirement, and foreign language), while the last topic (babies) generated almost 75% of total ads recorded during our testing phase. 

By device, Siri users reported the most ads, with about 78% of reported ads served overall. The other 22% of reported ads were from Google Home users. Our Amazon Echo user reported no ads. 

Most importantly, the majority of participants for all topics reported no ads, and for all devices (except for the baby topic) reported no ads. 

Insights Derived from the Qualitative Research Study about Google, Apple, and Amazon SmartSpeakers

So what have we discovered? Well, we don’t have enough data to support the theory that our devices are always listening and that that data is sold to advertisers. At least, not our SmartSpeakers at home. In every test we conducted, the majority of participants reported no ads during our testing period. 

That being said, the baby topic really came up big for us, didn’t it? It had dramatically higher ads than the three other topics, and interestingly, mostly for Siri users. So what does that mean? 

Well, put simply, it could mean that Apple sells user data that it collects via Siri. Some have long suspected that this was the case. Unfortunately though, it isn’t that simple. An overwhelming majority of the baby-related ads were served to just one participant, meaning that participant could have any other “digital baggage” that advertisers have been tracking. 

On top of that, it might not even be Siri that’s taking information from the phone’s recording devices. Many applications contain fine print in their terms of use that can be backdoors for advertising data to be collected.

Our study could (and should) easily be replicated on a larger scale over a longer period of time to find results that would be more indicative of the truth. Unfortunately, though, until companies that are dealing with this data are more transparent, we’ll just have to take their word for it. 

If you’re worried about your phone selling data from you, though, you can get ahead of it by deactivating “Hey Siri,” “Ok Google,” or whatever your equivalent is on your phone. Familiarize yourself with the terms of service on the applications installed to your phone, and be more aware of the data you’re putting out onto the web!

About the Authors: 

Leah Ellsworth is getting a Master’s degree in Advertising at the University of Texas. She graduates in May 2022 and can be reached at leah.ellsworth@texas.edu.

Lindsey Jamieson is getting a Master’s degree in Advertising at the University of Texas. She graduates in December of 2021 and can be reached at lindsey.jamieson@austin.utexas.edu.

Andrea Lloyd is getting a Master’s degree in Advertising at the University of Texas.  She graduates in May of 2021 and can be reached at aclloyd@utexas.edu.

Alison Gibson is getting a Master’s degree in Advertising Analytics and Insights at the University of Texas. She graduates in December of 2021 and can be reached at Alison.Gibson@utexas.edu.

Kellen Arnold is pursuing a Master’s degree in Advertising at the University of Texas and will graduate in December 2020. He can be reached at kellen.m.arnold@gmail.com.

Cara Bristow is getting a Master’s degree in Architecture at the University of Texas. She graduates in May of 2021 and can be reached at caranbristow@utexas.edu.

Alexandra Pease is pursuing a Master’s degree in Architecture at the University of Texas after receiving her Bachelor of Design degree from the University of Florida in 2019. She’ll be graduating May of 2021 and can be reached at alipease97@gmail.com.

Katherine Smithroat is pursuing a Master’s degree in Information Studies at the University of Texas after receiving her Bachelor of Arts and Sciences degree from the University of Texas in 2018. She’ll be graduating August of 2020 and can be reached at ksmithroat@utexas.edu.