Key Takeaways

  • Adam Foroughi dismisses the myth that phones eavesdrop on conversations to serve ads, explaining that cross-site search footprints and browsing history already create fast predictive matching.
  • Precise location tracking is too technically heavy and unnecessary for ad networks, with AppLovin avoiding location data entirely to run ad rendering.
  • When Apple implemented restrictive privacy controls on iOS, ad networks had to group users into broad cohorts, which degraded initial ad quality and triggered user complaints about spam.
  • Advanced deep learning models adapted to aggregated cohort boundaries once Apple established clear regulatory guardrails.
  • High-performing digital advertising expands the wider economy by driving product discovery without needing bottom-funnel search intent.

The Eavesdropping Myth and Digital Footprints

Everyone has experienced it: you mention a pair of running shoes over lunch, open your phone ten minutes later, and see an ad for that exact brand. The immediate assumption is that your microphone is secretly recording your conversations.

AppLovin CEO Adam Foroughi says that theory is pure fiction. Advertising networks do not record audio, nor do they track precise physical location. As Foroughi explains: “I think you've done other actions that are trackable like do a search, browse a website, do a product search and you don't realize it and then you say something related to and you start seeing ads that are relevant. So, it's not like the mic is on or there's an app that has actually taken space.”

Capturing, streaming, and processing continuous audio streams across hundreds of millions of devices would crush battery life, eat bandwidth, and trigger massive data overhead. Ad tech does not need audio because digital breadcrumbs are already thorough. When you browse three websites, tap an item in an app, or search an adjacent query, machine learning models connect the graph before you finish speaking.

The Cost of Heavy Location Data

Founders often assume granular GPS tracking is the fastest path to conversion. Foroughi points out that location data is both technically inefficient and commercially unnecessary for performance advertising.

“I don't think advertising companies can track location,” Foroughi notes. “So, we don't track location at all. It's a really heavy concept to track people's precise location to then run render an ad.”

Processing high-frequency latitude and longitude coordinates introduces latency into real-time bidding auctions. Sub-second auction environments require speed and pattern matching, not heavy geospatial queries. AppLovin built a high-margin business by stripping away bloated data overhead and focusing purely on conversion intent and behavioral models.

The Apple Privacy Paradox

When Apple cracked down on cross-app tracking on iOS, critics predicted mobile advertising would collapse. The real outcome was messier and counterintuitive.

“Look in any of these spaces you want the regulations to be clear,” Foroughi explains. “So once they're clear technology can deal with them and so if you could precisely target a user 5 years ago on iOS and today someone says I don't want you to precisely target me you group them in a bunch and you serve them a worse advertisement.”

Instead of delighting consumers, stripping away precision targeting made the user experience worse. Foroughi points out the immediate user backlash: “Now, the funny outcome of that is we'll get a lot of complaints after that change that Apple made from users that say, 'Serve me more relevant ads. You're showing me a bunch of spam.'”

Once the rules of the road became static, AppLovin trained deep learning models to predict intent from cohort-level signals. As Foroughi argues, effective digital advertising expands the wider economy: “There's a big part of GDP that's now coming from this digital ad economy. The better these technologies get, faster GDP growth.”

What to Do With This

Audit your acquisition analytics this week and purge low-value, high-latency data collection like precise GPS coordinates or redundant device pings. Rebuild your audience segmentation around aggregated behavioral triggers and cross-session search footprints rather than brittle personal identifiers. When platform privacy rules shift, stop lobbying for exceptions and retrain your predictive models on cohort-level patterns.