It happened to many with an almost offensive precision. A chat on the sofa, the name of an electric bike thrown out there, a holiday mentioned without searching it on Google, then the screen that a few hours later returns that very announcement. The temptation is immediate: the telephone listens. The scene is perfect, it has the flavor of a domestic rehearsal, the kind that is told with the eyebrow raised and the cell phone placed face down on the table, as if turning it was enough to silence a spy.
The most disturbing explanation, however, is often also the least theatrical. The microphone always on it is a weak, expensive, noisy track from a technical point of view. The advertising machine, however, has learned to work better: it cross-references position, searches, history, purchases, installed apps, digital friends, sites visited, advertisers who upload customer lists, pixels scattered across the pages and small signals left with the thumb every day. The result can feel like telepathy. It looks much more like a detailed accounting of our habits.
The myth of the turned on microphone holds little water
The theory of the telephone recording conversations to serve personalized ads has enduring appeal, not least because it arises from credible everyday experiences. Yet technical analyzes have found little material to support it on a large scale. A group of researchers from Northeastern University looked further 17 thousand Android apps looking for suspicious microphone activations and hidden audio sending: the most relevant result concerned screenshots and screen recordings sent to third parties by some apps, while secret ambient listening for advertising remained without solid evidence.
Even the most material part of the matter weighs heavily. Recording and transmitting audio continuously would act like an always open call. Such surveillance would leave footprints too large to remain invisible for long. Companies then continue to deny the use of the microphone for advertising purposes. Half states that Facebook and Instagram only use the microphone when the user grants permission and activates a feature that requires it. Appleafter the $95 million settlement in the lawsuit related to accidental activations of Siri, reiterated that it had never sold Siri data or used that information for marketing profiles.
This still leaves a real gray area: voice assistants can activate by mistake, some apps ask for broad permissions, and each operating system remains a huge surface of delegated trust. Androidfor example, shows indicators when an app accesses your microphone or camera, precisely because those sensors are delicate enough to deserve visible signals. The question, however, changes shape: rather than an army of always open microphones, we have an ecosystem that thrives on permissions, tracking and inferences.
Targeted advertising arrives before your voice
Targeted advertising works because you don’t need to hear something said over dinner. Google explains that personalized ads can be based on your activity, preferences, recent searches, content watched, approximate location, partner sites and apps. Applewith App Tracking Transparency, describes tracking as the ability for an app to follow a user’s activity across other companies’ apps and sites for advertising purposes or for sharing with data brokers. This definition already says a lot.
An afternoon with a friend is enough to build an association. Two phones in the same place, maybe for hours. One of the two is looking for hotels in the Maldives, the other scrolls Instagram without thinking. The advertising system sees proximity, similar interests, overlapping social networks, indirect contacts, behavior of “similar” people. After a while an advert for flights or resorts also appears on the phone and the search has only been breathed in from afar. From the outside it seems like I’m listening. From the inside it is statistics applied to an enormous amount of data.
The psychological part also comes in here. There frequency illusionoften called the Baader-Meinhof phenomenon, describes that feeling where, as soon as we notice an idea or object, we start seeing it everywhere. Selective attention does the rest: we scroll through hundreds of ads every day without really registering them, then one coincides with a conversation we just had and becomes an emotional test. The brain keeps that case, discards all the others.
The uncomfortable part is here: the system manages to seem personal because it is already personal enough. The phone knows the places we go through, the apps we open in the morning, the pages we stay on for a few extra seconds, the purchases left in the cart, the newsletters read, the videos interrupted, the contacts with whom we share physical and digital spaces. The microphone almost becomes a narrative accessory. The most effective surveillance comes from the screen.
Exceptions exist, and they say something different
There are cases in which the microphone has actually entered advertising or tracking mechanisms, even if with different logics from listening to human conversations. In 2016 the Federal Trade Commissionthe US authority that monitors consumers and competition, sent warnings to some developers who used SilverPush code: that technology could recognize audio beacons inserted into television content, signals imperceptible to people and capable of being picked up by the smartphone to understand which commercials or programs were playing near the device.
More recent and louder is the story of Cox Media Group and its service called “Active Listening”. In 2024, promotional materials circulated in which the company claimed it could target ads using voice data collected from smart devices; Google removed it from its partner program after questions from journalists and Meta distanced itself.
In May 2026, a decisive detail arrived: the FTC accused Cox Media Group and two other companies of deceiving customers by falsely claiming that it could target ads based on conversations collected by devices and that it had consumer consent. According to the authority, that commercial promise was deceptive. Even this story, born as a confirmation of the nightmare of the microphone, has become a different photograph: the surveillance market even sells the fear of surveillance.
The real flaw lies in the data we leave looking at
The most concrete point concerns data brokers. In the United States the FTC has intervened several times against companies accused of selling or sharing sensitive location data, capable of revealing visits to clinics, places of worship, shelters for victims of domestic violence and other very delicate spaces. That data often comes from apps, advertisements, real-time auctions and information packages resold along difficult to read supply chains.
App research shows even more ordinary fragility. Privacy labels can be vague, incomplete, difficult to read; many applications integrate third-party code, advertising libraries, analytics tools. A study on app iOS And Android detected widespread tracking and sharing of identifiers even in sensitive categories, including products aimed at children.
The case discovered by Northeastern researchers remains instructive for this very reason. They were looking for secret microphones, they found apps capable of recording the screen or capturing screenshots and sending them to external services. It is a less cinematic violation, much more tangible: what we type, touch, look at and grant to apps can tell more than a sentence said out loud.
To protect yourself you need less folklore and more maintenance. Checking microphone permissions is still useful, of course. It is also worth turning off location access when unnecessary, limiting cross-app tracking, emptying advertising identifiers, reducing social logins used everywhere, denying automatic permissions to apps that ask for too much, reading at least the main labels on data collection. These actions are small only in appearance, because they touch the real fuel of predictive advertising.
The telephone listens less than we fear. It observes us, connects us, classifies us, compares us with similar people, follows us in the most banal steps of the day. The microphone is scary because it looks like a spy with ears. The screen, meanwhile, takes notes.
