You Mention a Product Once, and the Ad Appears. Is Your Phone Listening?

It’s happened to you. You’re having tea with a friend, talking about trekking shoes. Just talking, phone face-down on the table. That evening you open Instagram, and there they are. The exact shoes. Maybe the exact brand.

Your stomach drops a little. The only explanation that makes sense is the creepy one: the phone was listening.

It almost certainly wasn’t. But the real explanation isn’t comforting. In some ways it’s stranger: the advertising industry knows you well enough that it doesn’t need your microphone.

The verdict

First, the short answer: no

People have genuinely checked. In 2018, researchers at Northeastern University tested over 17,000 popular apps, watching every byte each one sent out, waiting for secret audio. They found none. Not one app quietly recorded conversations and shipped them to advertisers. (They did catch a few apps doing other shady things, like recording the screen, but no hot mics.)

There’s also a boring, practical argument. Constantly processing audio drains battery and data in ways both engineers and the phone’s own battery-usage screen would notice immediately. Both Apple and Android also show an indicator dot whenever any app touches the microphone. A conspiracy this size, involving thousands of companies, with this many ways to get caught, and after fifteen years nobody has produced solid technical evidence of it.

And the deciding reason is the least flattering one: listening would be expensive, illegal, and (this is the part that stings) unnecessary. The ad appeared through channels far cheaper than audio.

The trick

How the trick actually works

Here’s the plain-language version. That ad found you through some mix of these:

1. You were with your friend, and the system knows it.
Your phones sat on the same table, on the same Wi-Fi, in the same location. Ad systems treat “these two devices spend time together” as a relationship. If your friend searched for those shoes yesterday (and they probably did; that’s why they brought them up), showing you the same product is an obvious, cheap bet.

2. You already left a trail and forgot.
A glance at a shoe review three weeks ago. A trekking video you didn’t skip. A friend’s photo you lingered on. None of it felt like “shopping,” so you don’t remember it. But it was all recorded, and it all points the same direction.

3. People like you buy things like this.
This is the piece most people underestimate. Advertisers build “lookalike audiences”: find everyone who bought the shoes, then find everyone whose behaviour resembles theirs in age, place, interests, the accounts they follow, and the hours they’re online. If ten thousand people who look statistically like you bought trekking shoes this month, you were getting that ad anyway. The conversation didn’t trigger the ad. The same forces that led to the ad (season, your circle, your feeds) also led to the conversation.

4. The ad was probably there before. You just didn’t see it.
Psychologists call it the frequency illusion. You scroll past hundreds of ads a week and register almost none. After the conversation, “trekking shoes” is suddenly loaded in your mind, and nowthe ad jumps out. The ads for things you never discussed? Invisible. Nobody texts a friend saying “weird, I got an ad for something we never talked about.” The misses are forgotten; the one spooky hit becomes a story.

Under the hood

The technical version

For the readers who want the machinery, here’s what’s under the hood.

Your phone carries an advertising ID, a resettable serial number that apps can read and attach events to. Every app with an ad or analytics SDK inside it (most of them) reports what you do alongside that ID. Data brokers merge these streams, plus purchase data and location pings, into a profile, then link your phone, laptop, and TV into one identity using logins and shared IP addresses. That linking is called a device graph, and it’s how two phones on one home Wi-Fi end up seeing related ads.

Then, every time a page or feed with ad space loads, an auction runs. This is real-time bidding. In roughly a tenth of a second, the ad exchange broadcasts “user with these traits is looking at ad slot,” advertisers’ systems consult their models and bid, and the winner’s ad renders before the page finishes loading. The targeting models are trained on billions of these events. They don’t need to hear you say “trekking shoes.” They predicted weeks ago that you were the kind of person about to want them, with accuracy that feels like eavesdropping.

That’s the honest summary of the whole phenomenon: it’s not surveillance of your voice. It’s prediction from your behaviour, and prediction got good enough to feel like magic.

The controls

What you can do about it

You can’t opt out of prediction entirely, but you can blunt it. Reset or delete your advertising ID (both iOS and Android let you, under privacy settings; on iPhone it’s “Allow Apps to Request to Track,” which you can simply turn off). Audit which apps have microphone and location permission, and revoke what doesn’t obviously need it. Not because of secret listening, but because location is one of the strongest targeting signals there is. And in your Google and Meta account settings, turn off ad personalisation. You’ll see the same number of ads, just dumber ones. Some people find that reassuring. Others miss the relevant ads within a week. Both are reasonable.

The point of understanding this isn’t paranoia. It’s calibration. The scary story about the microphone distracts from the true story, which is quieter and bigger: an industry that models your behaviour well enough to finish your sentences. Once you see how it works, the spooky moments stop being spooky. They’re just statistics, wearing a good costume.

— Santam Kumai builds websites and digital tools, and explains technology in plain language. Got a “my phone is doing something weird” question? WhatsApp him; it’s really him replying.

← All writing