Is your phone listening to you for ads?
“My phone is listening to me.” I hear that quite often. You talk to someone about a product, and soon afterwards it turns up in your ads. I understand why that feels suspicious, but I'd look at other explanations first.
In Panoptispy, a 2018 Android study (opens in a new tab), researchers found no audio leaks during their tests. They also described limitations of those tests. That doesn't guarantee no app is listening today. A well-timed ad doesn't prove it either.
Why that ad suddenly stands out
After talking about a subject, you pay more attention to it. An ad you scrolled past yesterday might catch your eye today. That fits what Arnold Zwicky calls the frequency illusion (opens in a new tab): once you notice something, it seems to be everywhere.
Then there's the information services already collect. Google lists (opens in a new tab) search activity, interactions with ads, device information and location, among other things. Which data gets used depends partly on your settings. There's already quite a bit of context there.
What shared Wi-Fi can reveal
Devices on the same Wi-Fi often share a public IP address. The US regulator FTC described back in 2017 (opens in a new tab) how advertising companies can use that to link devices to the same household. Location and other signals can help improve that estimate.
Activity on one device can then influence ads on another. If an ad network links you to the same household and also records the product visit, that could help explain why you see ads for something someone else viewed. The UK's CMA describes this kind of linking too (opens in a new tab). Briefly sharing Wi-Fi doesn't mean every search gets passed along.
When your speaker thinks you're talking to it
With smart speakers, the microphone is very much part of the job. An Echo with its microphone enabled listens locally for its wake word, Amazon explains (opens in a new tab). Once activated, it can send audio to the cloud. Researchers have also studied accidental activations (opens in a new tab): a device can think you're addressing it.
South Park played with that in 2017. Commands in the episode also activated viewers' speakers, causing unwanted alarms and strange shopping lists (opens in a new tab). Some viewers unplugged their devices.
AI adds more context
Meta announced in 2025 (opens in a new tab) that it would use AI conversations to personalise content and ads, rolling this out in most regions. That doesn't automatically cover every Belgian account. I suspect the extra context will make ads feel uncannily accurate more often.
ChatGPT can also show ads on Free and Go, depending on the region. OpenAI says (opens in a new tab) the current chat can help determine which ad you see. Personalisation using previous chats and memory has separate settings and is initially unavailable in the EEA, which includes Belgium. Advertisers don't get to see your chats.
Conversations in personal ChatGPT accounts can also be used for model training (opens in a new tab), including on paid plans. You can turn this off for new chats. Training and ad personalisation have separate controls; ChatGPT Business doesn't use business data for training by default (opens in a new tab).
For sensitive business work, I'd avoid free AI tools with ads. Don't casually paste in customer details, contracts or internal plans. Check the terms for data use and talk to IT first, even if you pay for the tool.
I also offer AI consultancy for businesses. Together, we look at which tools suit your work and what information you can share with them, alongside your IT team. Send me an email if you'd like to talk it through.