Smart speakers raise a genuine, common question worth answering accurately rather than dismissively or with exaggerated alarm.
How wake-word detection actually works
Smart speakers do continuously process nearby audio locally on the device itself, listening specifically for a wake word ("Alexa," "Hey Google," and similar). Under normal, intended operation, only after that wake word is detected does the device begin actively recording and sending audio to the cloud for processing. This is the honest technical answer to "is it always listening" — the device is always listening for the wake word specifically, using on-device processing, not continuously streaming everything to a company's servers.
The more accurate, non-hypothetical concern is different: wake-word detection isn't perfect, and devices do sometimes trigger accidentally on a similar-sounding phrase, sending audio that wasn't actually meant to activate the device. Several major manufacturers have also, in the past, had human contractors review a portion of recorded voice interactions for quality improvement purposes — a real, reported practice, not a hypothetical one, though companies have adjusted these programs and disclosure practices over time in response to public concern.
What you can actually control
Most smart speakers offer a physical microphone mute button or switch that provides a genuine hardware-level cutoff, more reliable than a purely software-based setting, and most platforms let you review and delete stored voice recordings through an account settings page. Checking and periodically clearing this history, and understanding what a specific platform's actual data retention and human-review policy currently is, gives a more grounded picture of a specific device's privacy posture than general assumptions in either direction.
Smart home hub compatibility varies by ecosystem
Beyond voice assistant functions, many smart speakers double as a hub connecting smart home devices — lights, locks, thermostats, plugs. Compatibility between a specific speaker's ecosystem and specific smart home device brands isn't universal; checking that devices you already own, or plan to buy, are actually compatible with a given speaker's platform avoids ending up with smart home gadgets that can't actually be controlled through the voice assistant you chose.
Multi-room audio requires matching ecosystems
Playing synced audio across multiple speakers in different rooms generally requires speakers from the same platform or ecosystem — mixing brands often breaks this specific feature, even if each individual speaker otherwise works fine on its own. Worth confirming before assuming any two smart speakers can be grouped together for whole-home audio.
The one thing people forget
Check what happens to smart home device control if the internet goes down, since many smart speaker-connected devices route commands through the cloud rather than working purely on a local network. A power or internet outage can mean losing voice control over lights or locks specifically when it might matter most, which is worth knowing rather than assuming smart home control always has a reliable local fallback.