I still don't think companies serve you ads based on your microphone
Claims that smartphones secretly listen to conversations to target ads are scrutinized against technical realities, legal risks, and the lack of hard evidence such as packet captures or credible leaks. Many commenters argue that existing data sources (search history, location, social and household associations, smart TVs, data brokers) already allow eerily precise targeting without live audio surveillance, and that large platforms are too exposed to risk running a covert microphone-based system at scale. Others counter with personal anecdotes, marketing pitch decks, and lawsuits like Apple’s Siri settlement to argue that accidental capture, opaque “QA” uses, or smaller players and embedded SDKs could still be exploiting ambient audio in limited or experimental ways.
Whether microphones are used for ad targeting
- Many commenters argue large platforms are not secretly using always‑on microphone data for ad targeting.
- Others are convinced it happens, citing repeated personal anecdotes where niche spoken topics quickly show up in ads or even postal spam.
- Some suggest a middle ground: if it happens, it’s more likely rare, experimental, or via smaller players/SDKs than a pervasive, coordinated system across major OS vendors and ad networks.
Technical feasibility and constraints
- Several adtech and big‑company insiders say privacy processes, legal risk, and organizational controls would make a secret, large‑scale audio‑to‑ads pipeline extremely hard to deploy and hide.
- Others counter that on‑device keyword spotting is easy and cheap: wake‑word chips, song ID features, and modern low‑power audio ML show it’s technically feasible to listen for many keywords and upload small tags instead of audio.
- Debate over power use and bandwidth: some say continuous rich speech recognition would kill batteries and be observable; others present rough power budgets indicating limited keyword spotting plus selective processing is plausible.
Alternative explanations for “creepy” targeting
- Frequent suggestions:
- You or a friend recently searched for the topic; co‑location, shared IPs, and social graphs propagate targeting.
- Location and demographics (e.g., living on an island → kayak ads; public housing → “aspirational” luxury brands).
- Smart TVs and other devices doing automatic content recognition or voice‑assistant logging.
- Classic cognitive effects: confirmation bias, Baader–Meinhof/frequency illusion, misremembered browsing, and selection bias in which stories get told.
- Some argue that these non‑audio signals are powerful enough that users feel like their phones must be listening.
Evidence cited: lawsuits, decks, and leaks
- Apple’s Siri lawsuit and $95M settlement: commenters note it showed inadvertent recording around false wake‑words and problematic QA use, but no proof of ad targeting.
- “Active listening” pitch decks (e.g., from Cox Media or similar): widely debated; many see them as vague, aspirational sales material, not confirmation of large‑scale smartphone eavesdropping.
- No one in the thread can point to a clear packet capture, reproducible experiment, or credible insider leak proving continuous microphone‑based ad targeting.
Privacy attitudes and implications
- Several worry that fixation on the “phone is listening” myth distracts from documented abuses: extensive location tracking, data brokers, cross‑device profiling, and smart‑TV surveillance.
- Others argue that, given incentives and history (surveillance capitalism, past scandals), it’s rational to remain suspicious even without hard proof.
- Some note broad public fatalism: many people believe they’re being surveilled yet change no behavior, which weakens pressure for real privacy reforms.