A presence sensor is the device that is supposed to notice you are still in the room after you stop waving your arms. Ordinary motion sensors, the kind that have run hallway lights for years, look for a change. Sit down to read and they decide the room is empty. In 2026 a wave of small millimeter-wave sensors, usually shortened to mmWave, promises a finer trick: they notice breathing-scale movement and can hold a light, a fan, or a thermostat while you are quiet. That promise is real enough to shop for, and narrow enough to waste money on if you buy a gadget when you needed a lamp switch.
The radio is not a camera, and that is the first reason people choose it. A decent presence sensor does not record a picture of the room. It reports occupied or not, and on the better units a coarse sense of where in the room the movement is. It can still reveal that someone is home, and a sloppy automation can announce that fact to a phone, a speaker, or a log you did not mean to share. Buy it as a lighting tool first. Treat any cloud account as optional. A sensor that cannot light a room unless a server agrees has failed the basic test, because the person on the couch is local and the internet is not required to see them.
What it can tell, and what it will invent
These sensors are good at "someone is here" in a bedroom, a study, or a bathroom where a motion sensor gives up. They are bad at knowing who. They do not distinguish you from a partner, a teenager, or a large dog if the dog is in the zone. They can be fooled by a fan, a curtain in a draft, a robot vacuum, or a pipe moving with hot water. The setup that works is a room with a job: keep the desk lamp on while the chair is occupied, turn the bath fan off ten minutes after the room is truly empty, stop heating a guest room that has been vacant all afternoon. The setup that fails is a whole-house brain that tries to infer a family's mood from one anonymous blob.
Range and angle matter more than the slogan. A sensor aimed at a hallway from a corner may see the hallway and also the room through the door, then leave a light on for a person who only walked past. A sensor behind a television, or pointed at a window, may chatter. Mount it where the manual says it can see the place people actually sit, not where the cable happens to reach. Ceiling mounts and wall mounts see different shapes. If you rent, prefer a unit that stands on a shelf or screws into a place you can patch. A sensor that requires a new electrical box is a project, not an accessory.
Zones, still bodies, and the pet on the sofa
The feature worth paying for in 2026 is a zone you can shrink. You want the reading chair included and the doorway excluded. You want the ceiling fan either outside the logic or filtered so it does not count as a person. Units that offer this only inside a phone app are usable if the result is stored on the sensor or on a local hub. Units that forget the zone when the app is uninstalled will be reconfigured every time you change phones. Do the boring test: sit still for ten minutes with a book. Then leave for ten minutes. The light should stay, then go. If it drops while you are on the page, the sensor is still a motion sensor with a higher price.
Pets are the honest complication. A sensor sensitive enough to see a person breathing may also see a cat at the foot of the bed. Some firmware lets you ignore targets below a size or outside a height band. That filter is imperfect, and a dog who sleeps on the sofa will keep the living room "occupied" all evening, which may be what you want or exactly what you do not. Decide before you automate the heating. Holding a setpoint for an animal can be kind. Doing it by accident, in every room, is how a sensor raises a bill and gets unplugged.
Local control, Matter, and the hub you already have
Ask where the decision is made. The better pattern is a sensor that talks to a hub you already run, over a local protocol, and then that hub switches a lamp or a relay. Matter, Zigbee, Z-Wave, and a plain contact wired to a local controller can all be respectable if the path stays in the house. A sensor that only speaks through a vendor cloud will fail open or fail closed when the account does. Know which. A light that sticks on all night is annoying. A light that goes off while you are in a windowless room is worse. Prefer the failure that leaves you able to use a normal switch.
Also ask what the sensor does in a power cut and after. Battery units are easy to place and easy to forget until the light stops behaving. Mains-powered units are honest about being installed. Neither is better for every room. A bathroom may justify a wired sensor. A rental study may justify a battery unit you can take when you leave. Confirm that a battery warning is visible without opening a cloud dashboard you check twice a year. The automation should degrade into an ordinary room, not into a puzzle.
Who it is for
Buy a presence sensor if a specific room keeps going dark or shutting down while you are quietly in it, and if you are willing to spend an hour aiming it. Buy the local one. Skip it if what you wanted was a camera, a security system, or a way to track a person through the house. Those are different products with different duties to the people who live there and the people who visit. A presence bit is a small fact. It should stay small.
Skip it also if the room already has a switch everyone understands and the "problem" is only that an app would be neat. Neat is how drawers fill up with radios. The 2026 sensor that earns its place is the one you stop noticing: the lamp stays on while you read, it goes off after you leave, the fan does not run until morning, and none of that required a subscription. If the product page leads with a cloud plan and a picture of a family being observed, close it. You were shopping for a light that understands a still person. That is a modest goal, and it is enough.
Frequently asked questions
Is an mmWave sensor a camera?
No. It does not capture an image you can look at. It reports presence, and sometimes a rough position. It can still reveal that a room is occupied. Do not send that fact to a service you would not tell directly.
Will it keep the lights on while I sit still?
That is the point of a good one, and it is the test you should run before the return window closes. Sit and read for several minutes. If the room drops to "empty" while you are there, aim it again or return it. Do not invent a habit of waving at the ceiling.
Can it tell people apart?
No. Treat every claim of identifying a person as a different, more invasive product. A presence sensor should not be your lock, your attendance system, or your way of watching a household. It knows movement in a zone. That is all.
Do I need a hub?
You need some local way to turn the signal into an action, unless you are only using the sensor's own relay to feed a fixture. A hub you already trust is better than a new cloud account. The sensor should still be understandable if that company's app disappears.
Will pets set it off?
Often, yes. Filters help and do not always succeed. If a pet sleeps in the room you want to automate, test with the animal there before you let the sensor run the heating overnight. An occupied sofa is a true reading, even when the occupant has four legs.