Your phone isn’t secretly listening to you for ad targeting
The notion that smartphone owners’ ambient conversations are monitored for ad targeting foments into a public furor every so often. Most recently, 404 Media published a report revealing that a subsidiary of the Cox Media Group claims to be able to use voice data collected surreptitiously from smartphones to target ads on various channels; the report was picked up by outlets such as The Daily Mail, New York Post, and Sky News Australia.
The relevant slide from Cox Media Group’s recent sales deck is above, but its assertion of capability is not new; it surfaced in December 2023 from a similar sales pitch deck, where the tool was given the name “Active Listening,” for which the company declared:
It’s True. Your Devices Are Listening to You. With Active Listening, CMG can now use voice data to target your advertising to the EXACT people you are looking for.
The revelation catalyzed a news cycle at the time, and the Cox Media Group subsequently scrubbed its website of all mentions of the tool. This particular advertising agency’s ability to deliver this type of functionality is suspect; the slide above is exceptionally vague and doesn’t explain how AI, voice data, and ads targeting feasibly fit together. But the idea that smartphones parse conversations for specific keywords to use in targeting ads didn’t start with this agency. Mark Zuckerberg was asked pointedly in a Senate hearing by Senator Gary Peters (MI) in 2018 if its apps use smartphone microphones to collect ambient voice data to target ads — he responded, directly, that the company doesn’t. The relevant portion of the exchange is below:
PETERS: And — and I think an example of this is something that I’ve been hearing a lot from folks that have been coming up to me and talking about really, kind of the experience they’ve had, where they’re having a conversation with friends.
Not on the phone, just talking. And then they see ads popping up fairly quickly on their Facebook. So I’ve heard constituents fear that Facebook is mining audio from their mobile devices for the purpose of ad targeting. Which I think speaks to this lack of trust that we’re seeing here, but — and I understand there’s some technical issues and logistical issues for that to happen.
But for the record, I think it’s clear — see, I hear it all the time, including from my own staff. Yes or no, does Facebook use audio obtained from mobile devices to enrich personal information about its users?
ZUCKERBERG: No … Well, senator, let me be — let me be clear on this. So you’re — you’re talking about this conspiracy theory that gets passed around that we listen to what’s going on, on your microphone and use that for ads.
PETERS: Right.
ZUCKERBERG: We don’t do that.
It seems unlikely that the CEO of a publicly traded company would blatantly lie under oath during a Congressional testimony. But if Meta doesn’t furtively harvest voice data from smartphones for ads targeting — and it is perhaps the company most associated with the hypothetical practice — then what would need to be true for any other company to do so? Answering that question requires surveying the roadblocks that exist for the practice:
- Both iOS and Android require an app to receive user permission to access the device microphone. When permission is granted, both operating systems allow apps to access the microphone when the app is in the background, although Android limits the amount of time and use cases for which an app can do so;
- Both iOS and Android indicate when the microphone is in use through lights on the screen. When the indicator is on, users on both platforms can detect which apps are using the microphone with a limited amount of investigation;
- Both iOS and Android allow users to audit data transmission per app. Harvesting voice data continuously, or even just when certain keywords are triggered, and transmitting that data back to some central server would utilize a large volume of data, which could raise the suspicion of users.
Beyond these restrictions, for a company to be able to harvest this data, and enough of this data for it to be compelling to advertisers, the company would need either a large distribution footprint — its app installed on many devices — or an ability to purchase this data from a third party that does. Any company that professes this functionality would need to credibly prove access to the underlying data. Certainly, many companies — like social media platforms and game developers — can claim this. But how many advertising agencies can?
Further, more practically, anyone selling a tool like Active Listening would need to convincingly explain why voice data parsed by keywords is useful for ads targeting. The examples shared in the most recent version of the deck (above) are either of dubious value (“Do I see mold on the ceiling?”) or simply seem unnatural in everyday conversation (“The car lease ends in a month- we need a plan”). Why would these keywords as stated in conversation be more indicative of intent than when included in searches? And how likely are these keywords to be used in a commercially relevant way in conversation versus incidentally (“A car lease is a terrible financial decision”)? The same can obviously be said of keywords targeted in search, but an advertiser has direct control of those.
Advertising agencies often overstate their capabilities. Perhaps the most infamous example of this is Cambridge Analytica, which claimed to be able to target potential voters based on psychographic profiles but in reality was determined by the UK’s ICO to have used “well recognised processes using commonly available technology” which raised “concern internally about the external messaging when set against the reality of their processing.” With respect to voice data collection, CBS News partnered with the Electronic Frontier Foundation to monitor outgoing traffic to potential audio data harvesters when commercially valuable keywords were uttered into a smartphone and didn’t observe any. The BBC engaged in a similar exercise and reached the same conclusion.
More realistic than any advertising agency possessing the capacity to harvest (or buy) tremendous volumes of audio data — and to be able to parse that data for relevant commercial intent — is that this practice exists as a perennial fallacy animated by broad, warranted digital advertising skepticism. Anyone looking for reasons to be critical of the data collection and utilization undertaken in digital advertising is spoiled for choice. But the idea that advertising platforms — and especially boutique advertising agencies — can harvest audio data from smartphones to use for ads targeting is almost certainly a delirious myth.
Comments: