TL;DR
- The claim: many users believe Facebook and Instagram listen through the phone microphone and show ads based on spoken conversations.
- The evidence: Meta has publicly denied using the microphone for ads since 2016, and no independent research has shown that it does.
- What it uses instead: on-platform behavior, location, the websites and apps that share data through the Meta Pixel and similar tools, life events, and the behavior of people with similar profiles.
- The precedent: the Cambridge Analytica scandal showed in 2018 that likes and quiz answers were enough to profile up to 87 million people. Meta later settled for $725 million in the US and A$50 million in Australia, without admitting wrongdoing.
- Why it feels like listening: cross-site tracking, lookalike targeting and confirmation bias.
Contents
- Where the listening theory comes from
- What Meta says about the microphone
- Cambridge Analytica: profiling without listening
- The signals behind ad targeting
- How retargeting works
- Why ads feel like mind reading
- FAQ
- Vyara’s take
A user mentions a product in conversation and sees an ad for it the next day. Stories like this are among the most widespread beliefs about social media. The more likely explanation is less dramatic and, in some ways, more revealing: the ad system rarely needs to hear anything.
Where the listening theory comes from
The theory rests on coincidences that feel too precise to be chance. Microphone permissions add to it: Facebook, Instagram and Messenger request microphone access for voice messages, calls, video and stories, so users know the apps technically can record.
One widely shared example, a company that supposedly knew a woman was pregnant before she did, is often attributed to Facebook. The original story is about the US retailer Target, which The New York Times reported in 2012 had built a model that predicted pregnancy from purchase patterns. It involved shopping data, not a microphone.
What Meta says about the microphone
In June 2016 Facebook published a statement saying it does not use the phone’s microphone to inform ads or to change what people see in News Feed, and that it only accesses the microphone when users have given permission and are actively using a feature that requires audio.
| Question | Status |
|---|---|
| Can the apps access the microphone? | Yes, with user permission, for audio features |
| Does Meta say it uses audio for ads? | No, it has denied this publicly since 2016 |
| Is there independent proof of ad listening? | No published research has demonstrated it |
| Can users check and revoke access? | Yes, in the phone’s privacy settings |
Cambridge Analytica: profiling without listening
In 2018 it emerged that a personality quiz app had collected data from Facebook users and their friends, and that the data had been passed to the political consulting firm Cambridge Analytica, which used it to build psychological profiles for political campaigns.
| United States | Australia | |
|---|---|---|
| People potentially affected | Up to 87 million worldwide | About 311,000 Australian users, according to the regulator |
| Settlement | $725 million, class action, agreed in December 2022 | A$50 million, enforceable undertaking accepted by the OAIC in December 2024 |
| Admission of wrongdoing | None | None, given without admission of liability |
The case is often cited in the listening debate for the opposite reason: it showed that behavioral data alone, such as likes and quiz answers, was enough to profile people at scale.
The signals behind ad targeting
| Signal | Example | Source |
|---|---|---|
| On-platform activity | Likes, comments, shares, saves, time spent on a post | Facebook and Instagram |
| Interests | Groups joined, pages followed, videos watched | Facebook and Instagram |
| Location | Home area, places visited, travel | Device and app settings |
| Off-platform activity | Products viewed, items added to cart | Websites and apps using Meta tools |
| Life events | Engagement, moving house, new job | Profile information and inferred behavior |
| Social graph | What friends and close contacts engage with | Facebook and Instagram |
Advertisers combine these signals with lookalike audiences, where Meta finds people who resemble an advertiser’s existing customers. An ad may therefore reach a user not because of anything the user did, but because people with similar profiles behaved a certain way.
There are limits. In January 2022 Meta removed thousands of detailed targeting options linked to sensitive topics such as health, race and ethnicity, political affiliation, religion and sexual orientation. Advertisers also see aggregated audiences, not individual profiles.
How retargeting works
The Meta Pixel is a piece of code that website owners add to their sites. It reports actions such as page views, add-to-cart events and purchases back to Meta.
| Step | What happens |
|---|---|
| 1 | A user visits an online pet store that uses the Meta Pixel |
| 2 | The user adds dog food to the cart but does not complete the purchase |
| 3 | The store runs a campaign aimed at people who abandoned their cart |
| 4 | The user sees the product repeatedly on Facebook and Instagram |
Combined with social signals, such as several close friends engaging with travel content about Greece after the user visited a holiday website, the result can look uncannily like the platform overheard a conversation.
Why ads feel like mind reading
- Confirmation bias. People remember the rare ad that matched a conversation and forget the many that did not.
- Shared causes. The conversation and the ad often have the same origin, such as a search, a store visit or a friend’s recommendation.
- Proxy signals. Friends, family and people nearby generate data that can reach a user’s feed.
- Timing. Retargeting campaigns often run within hours of a website visit, which makes the match feel immediate.
FAQ
Does Facebook listen to my conversations for ads?
Meta says it does not, and no independent research has shown that it does. The available evidence points to behavioral, location and cross-site data as the source of targeting.
Why did I see an ad for something I only talked about?
The most common explanations are a related search or website visit, activity by friends or people nearby, a lookalike audience, or confirmation bias. Users can tap the menu on an ad and choose Why am I seeing this ad to see the main factors.
What is the Meta Pixel?
A tracking code that websites install to measure ad results and build audiences. It reports actions such as viewing a product or adding it to a cart back to Meta.
Can advertisers see who I am?
No. Advertisers choose audiences and see aggregated results, such as reach and conversions, but not the identities of individual users.
Can I limit the data used for ads?
Yes. Meta’s Accounts Center includes Ad preferences and a setting for activity off Meta technologies, where users can review and disconnect data shared by other websites and apps. Phone settings control microphone access and app tracking.
What did Meta pay for Cambridge Analytica?
Meta agreed to pay $725 million to settle a US class action and A$50 million in Australia under an undertaking with the privacy regulator. In both cases it did not admit wrongdoing.
Vyara’s take
“Facebook doesn’t need to listen to you. Your behavior already tells the algorithm more than a microphone could: where you go, what you stop scrolling at, what you put in a cart and what your friends care about. As a marketer, that’s exactly why these ads work. As a user, the answer isn’t panic, it’s knowing which settings exist and using them.”
Further reading: Is Facebook Listening to You? What Really Drives Your Ads – Vyara Stefcheva’s practical guide on stefcheva.com.
Sources: Facebook Newsroom on microphone use · NPR on the US settlement · MinterEllison on the Australian settlement · NBC News on sensitive targeting options · The New York Times on retail predictions · Meta Pixel overview
About the Author
👉Want guest posting?🪙Subscribe to the @egorithms Telegram