The situation is familiar to almost everyone. You are discussing something with a friend that you have never searched for on the internet, and a few hours later you see an advertisement for it in your feed. The conclusion suggests itself, and according to surveys, approximately half of people are convinced that their phone listens to them for advertising. Research does not confirm this version. But the real explanation turns out to be not reassuring, but rather the opposite, since the actual mechanism works more precisely than any eavesdropping and is much harder to get around.

What the Checks Showed

The best-known study on this topic is the work of a team from Northeastern University led by David Choffnes, in which they tracked the network traffic of approximately nine thousand Android applications. They did not find hidden audio recording and its transmission in any case, including the applications that were most often suspected. At the same time, the researchers discovered something else, and it turned out to be unexpected. Some applications took hidden screenshots and sent them to third-party companies, and one even recorded video of the screen. A separate experiment in 2020, in which television audio was played to smart speakers for one hundred and thirty-four hours, also found no signs of continuous recording. A broader analysis covering more than seventeen thousand applications also found no cases of unauthorized microphone activation.

Why This Makes No Technical or Economic Sense

The microphone really does work constantly, but only to recognize the activation word, and a separate energy-efficient chip on the device itself does this. Audio is sent to servers only after the activation word has been triggered. Continuous transmission of audio to servers would require a huge amount of traffic, noticeably drain the battery, and would be immediately detected by researchers analyzing the network activity of applications. In addition, both mobile systems have already been blocking background access to the microphone without explicit permission for several years, and an indicator of an active microphone is shown on the screen. But the main argument is not technical, but economic. Behavioral models predict your intentions so accurately that sound simply would not provide advertisers with additional value, while the costs and risks would be enormous.

Then Where Does the Feeling Come From

The psychological part consists of the frequency illusion, also known as the Baader-Meinhof phenomenon, and confirmation bias. You notice the advertisement that coincided with the conversation, and you do not notice the hundreds of advertisements that did not coincide with anything. One striking coincidence is remembered, while a thousand ordinary impressions disappear from memory without a trace. The technical part is that advertising systems know enough about you for such coincidences to occur regularly without any microphone.

How They Actually Guess

The most interesting mechanism is called co-location. If you and the person you are talking to are connected to the same Wi-Fi network or are simply nearby, systems conclude that you are connected to each other and may show you advertising based on what the other person searched for.

That is, you did not search for this thing, but your friend searched for it an hour before the meeting, and that is why the advertisement looks as if it overheard your conversation. The same principle works for households where several people use a shared network and fall into the same target group. In addition, platforms use search history, behavior inside applications, location data, and the advertising identifier of your device. A separate and least transparent layer is data brokers, that is, companies that collect, buy, and sell detailed profiles of people. Advertising that looks as if it knows you often really does know you, because a broker sold your profile to a buyer with whom you have never had any dealings. Location data purchased from such brokers made it possible to identify visits to medical facilities and religious institutions, and this is precisely why the American regulator filed the relevant lawsuits. Another mechanism consists of matching customer lists, when a company uploads a list of contacts to an advertising system in encrypted form, and the platform finds matches among its users.

The Honest Complexity

It would be wrong to close the topic with a statement about the complete absence of problems, since several points deserve mention. Voice searches that you consciously make are processed as ordinary searches and may influence advertising. This is a documented mechanism, not hidden surveillance, but the confusion arises precisely because of it. In 2024, journalists discovered a presentation from a company that directly advertised to advertisers a function for active listening through device microphones. Google subsequently reported that it had stopped working with this advertiser, while Meta stated that it blocks targeting based on microphone data. That is, the offer itself existed, although evidence of it working at scale never appeared. Separately, advertising targeting should be distinguished from spyware. Targeted surveillance of a specific person through a program installed on their phone is a real phenomenon, and it has nothing to do with the question of advertising.

What to Do About It

Since the cause is not the microphone, turning off the microphone does not solve the problem, and you need to work with other mechanisms. Reset the device's advertising identifier and turn off personalized advertising in the system settings, since it is this identifier that connects your activity between applications. Review application permissions and remove access to the microphone, camera, and location from all programs that do not need it for their intended purpose. For those that do need it, choose the option to grant permission only while using the application. Turn off constant access to precise location, leaving approximate location where that is sufficient. Check the advertising preference settings in your Google and Meta accounts, where you can see which interests have been assigned to you and edit this list. Use a browser with tracker blocking, since the main part of tracking occurs not through the microphone, but through tracking your movement between websites. Limit what you publicly publish about yourself, since this is exactly the information that gets into data broker profiles. I wrote in a separate material about how to reduce your own digital footprint.

When You Should Really Be Concerned

If you suspect not advertising targeting, but surveillance specifically directed at you, the signs will be different. They include a sharp increase in data usage without explanation, rapid battery drain, the device heating up while idle, the appearance of applications that you did not install, and an active microphone or camera indicator at moments when you are not using anything.

One sign by itself proves nothing, but a combination of several that appeared suddenly is a reason to check the device.


Sources: research by Northeastern University led by David Choffnes, materials from Kaspersky, Security.org, 404 Media investigation into the advertising of an active listening function