We’ve recently had some odd conversations in the CNET offices.
Social media reporter Queenie Wong expressed a sudden interest in chainsaw sculptures. Laura Hautala, our privacy reporter, blurted out an assessment of the cost of skydiving lessons. (She says $200 is reasonable for an introductory class, not that she’d take one.) Abrar Al-Heeti, a general assignment reporter, described getting kicked out of a Casper mattress store in a suburb of Chicago after her brother’s kids went wild in it.
These offbeat exchanges weren’t random. They were part of an informal study CNET conducted to see if we could get Facebook to deliver ads based on discussions we had when our phones were in earshot. We chose the topics — some relatively unusual, others fairly common — to see if we could find indications the social network was aware of our conversations and using them to target ads. We didn’t find any.
The reason for our test: The long-running and hard-to-kill conspiracy theory that Facebook is listening to your conversations through the mic on your phone and then using this overheard dialogue to send you targeted ads. If you haven’t heard about this urban legend, there’s plenty of reading material online, including some lengthy threads on Reddit. The theory is so widespread that Facebook posted a formal denial three years ago and CEO Mark Zuckerberg denied it in testimony on Capitol Hill. Still, it just won’t die.
Our two-week test wasn’t meant to be exhaustive or scientific, though we did consult with security and systems experts to get some ideas on how to proceed. We weren’t trying to prove or disprove that Facebook was listening to our conversations. Our goal was more modest. We were just looking for any indication the conspiracy theory might merit a more rigorous investigation.
Two of the eight participants reported getting ads for Casper or Purple, the online mattress stores, after a long conversation about beds and box springs. But it turned out the reporter getting the Casper ad had visited a medical website after a backache caused by a poor night’s sleep. We suspect the research prompted the ad. The reporter who got the Purple ad had been receiving ads from the mattress maker before the test started. They simply continued. (A third participant got Purple ads before the conversation about mattresses, but not after. Go figure.) Discussions of more-specialized topics, such as an in-depth back-and-forth on the best chainsaw brands, prompted nothing.
Our test method
Here’s how we conducted our informal test, which involved four reporters in San Francisco, three in New York and one in Chicago.
We chose three topics to discuss in front of our phones. (On the advice of experts, we turned off geolocation and Wi-Fi services on our phones during the test so Facebook would have a harder time gaining our location information and showing us ads based on what people nearby were viewing online.) The topics were chainsaws, skydiving and mattresses.
For three days before the first conversation, the reporters monitored ads delivered while they used Facebook. What they saw was a mishmash of semipersonalized promotions, stores they shopped at or publications similar to the ones they read, and the usual stuff pushed on Facebook, such as Uber and Lyft.
On Friday, April 19, the reporters in San Francisco gathered to talk about chainsaws. In New York, reporters had these conversations independently. From Chicago, Abrar engaged in an amusing conversation by phone with an editor. The discussions, which lasted roughly 10 minutes, covered big-name brands, including Husqvarna, Stihl, Makita, DeWalt and a host of others. It inspired Queenie to express her profound appreciation of chainsaw carving.
Three days later, on April 22, the reporters gathered again to discuss skydiving. None of the reporters had jumped out of a plane. Had the reporters wanted to try parachuting, however, Facebook would’ve been of little help. No one got an ad recommending a skydive school.
The last conversation took place on Thursday, April 25. The topic this time was mattresses. We chose this topic because online mattress companies, such as Casper and Purple, often advertise on Facebook. If the social network was eavesdropping, we reasoned, this would be a wide-open opportunity for it to deliver ads based on our conversations.
Ian Sherr, a CNET editor, got an ad for Purple after the conversation. But he said he’d gotten the same ad before discussing the subject as part of the test. That led us to believe the ad wasn’t prompted by our discussion of mattresses, but rather by something Ian had previously looked at. Laura saw a Purple ad before our mattress discussion, but not after it.
Queenie got an ad from Casper after the discussion and was “kind of creeped out.” The ad even referenced the conspiracy theory, saying, “No, your phone’s not listening to you. We’re talented mind readers…” Then Queenie remembered she’d visited a Mayo Clinic page discussing relief for back pain after sleeping funny. We figure Facebook saw that research because it sees all the websites you visit if they participate in its pixel program or use its share, like and comment buttons.
To reiterate, our test wasn’t scientific and wasn’t designed to prove whether Facebook is or isn’t listening to you through your phone. But based on our results and discussions with security experts — and the fact that Facebook uses a myriad of other data points to base its targeted ads — we didn’t see anything that convinced us the social network is eavesdropping.
We’re also pretty sure that won’t kill the conspiracy theory that Facebook is eavesdropping via your phone, no matter how odd your conversation is.