Aircon motion sensor: what it detects and what it misses
A sensor that watches the room sounds like it can only help. Its worth depends on how many of your cooling hours happen in an empty room, and on whether the machine can tell an empty room from a still one.
By Team Snowflake | Updated 10 Aug 2026
What is the sensor actually detecting?
Two different devices are sold under one word, and they do not know the same things. One of them reports movement in front of the unit. The other measures temperature across the room and works out where a body is sitting. Which one is fitted decides everything the feature can do afterwards, and it decides which failures you will live with. The name on the brochure is no help, because both are sold as a sensor that knows whether you are there.
Daikin's operation manual states its position plainly. Intelligent eye is described there as the infrared sensor which detects the human movement, and the same manual divides the room into two areas so the unit knows which side of it moved. Movement is the entire input. The unit never learns that a person is present. It learns that something changed in area one or area two, and it counts the minutes since the last change. Everything the feature does afterwards is built on that single fact.
Mitsubishi Electric documents a different instrument. Its operation manual for a ceiling cassette states that the 3D i-see Sensor operates once approximately every 3 minutes to measure the floor temperature and detect persons in the room. That is a scan on a timer, not a trip wire. The unit reads a thermal picture of the floor and picks bodies out of it. That is how the same manual can discuss where an occupant sits, and how many people the room holds. A service manual for the wall-mounted range notes the sensor physically moving from side to side while it works.
The parts behind those two descriptions work on opposite principles. One component maker publishes both. Panasonic Industrial Devices states that pyroelectric sensors react to a change in intensity of infrared radiation. For its Grid-EYE thermopile array, the same company puts it another way. That sensor can detect temperature distribution in a two dimensional area. It identifies the number and position of people based on their body temperature. Change against picture is the whole difference. It is the first thing to settle about any unit that carries the feature.
The person who stops moving
Four of the makers quoted here admit the same failure in their own manuals. A person who stops moving can read as an empty room. Fujitsu General goes furthest and prints it as a warning instead of a note: do not use this function when there are only infants, children, elderly, or sick persons in the room. That is a manufacturer naming the households its own feature is wrong for, in the document that ships inside the box.
Daikin puts the warning next to the instructions. The manual states that the air conditioner may go into energy-saving operation even if there are people in the areas. It then names a reason. This may occur depending on the clothes the people are wearing if there are no movements of the people in the areas. Two conditions sit inside that sentence: what you are wearing, and whether you are moving. Neither is unusual in a room where somebody is reading, working at a laptop, or watching something to the end.
The thermal instrument carries the same limit, which is the part worth pausing on. Mitsubishi Electric lists the situations in which a person may not be detected, and a heat source that does not move for a long time is on that list. So is a person wearing heavy clothes whose skin is not exposed. A sensor that measures floor temperature ought in principle to see a motionless body, and the maker still names prolonged stillness as a miss. Paying for the more capable instrument narrows the problem. It does not remove it.
One condition on those lists belongs to this climate. Fujitsu General puts a number on it, naming a room at 30 °C or higher in summer as a case where the sensor may find nobody when somebody is there. Mitsubishi Electric lists a high room temperature in the same way. Panasonic Industrial Devices explains why from the component side, stating that during summer, when the difference between surface temperature of a human body and the background temperature is small, the sensitivity decreases. A Singapore room that has not been cooled yet routinely sits at that number or above it. Detection is weakest in the opening stretch of a cooling run.
What does the unit do once it decides nobody is there?
A clock does the work the sensor cannot. Daikin acts once no movement has been detected for more than 20 minutes, and Fujitsu General publishes the same wait. Hitachi shifts the setting at 20 minutes and shifts it again at 60. Mitsubishi Electric waits 60 minutes before its energy-saving step, and allows a separate shutdown timer set anywhere between 60 and 180 minutes. Those figures are not measurements of your household. They are how long each maker decided to wait before believing the silence.
What happens after the wait is where most of the industry agrees and one design does not. Daikin, Fujitsu General, Hitachi and Panasonic all respond by moving the target temperature, and none of those manuals describes the unit switching itself off. Panasonic adds a second lever, capping the fan at medium in cooling and dry while the room reads as absent. Mitsubishi Electric goes further again. It shifts the target, aims the vanes toward an occupant or away from one, counts how full the room is, and will stop the machine outright. Most designs nudge a number. One takes decisions.
Airflow steering changes what you feel before it changes what you pay, so it belongs apart from the energy argument. Daikin's manual describes the louvers adjusting so that the leftward or rightward airflow will not be directed to the people. Mitsubishi Electric turns the same idea into a choice. Its Indirect setting means an occupant will be less exposed to direct airflow. Its Direct setting means mainly the vicinity of an occupant will be air-conditioned. An Even setting learns the area where an occupant spends most of the time. Once the vane has moved, the stream behaves like any other cold jet. The coanda effect then decides how far it runs along the ceiling before it drops.
Neither approach is better, and neither manual claims it. They answer different questions. Mitsubishi Electric also prints a limit on the more capable one, noting that when more than a couple of people are in a room, the airflow control mode may work less effectively. That is worth knowing before the feature gets valued on a living room unit. What a buyer chooses between is a feature that adjusts a number quietly and a feature that will act on the room without being asked.
| Response once the room reads as empty | Where the makers document it | What you notice when the reading is wrong |
|---|---|---|
| Response once the room reads as emptySet temperature shifted | Where the makers document itDaikin: +2 °C in cooling after 20 minutes with no detection. Fujitsu General: up to +2 °C after about 20 minutes. Hitachi: two shifts, at 20 minutes and again at 60. Mitsubishi Electric: 2 °C after 60 minutes, plus 1 °C when counted occupancy drops | What you notice when the reading is wrongThe room drifts warm while the remote still displays the number you chose |
| Response once the room reads as emptyFan speed capped | Where the makers document itPanasonic: during human absence, maximum fan speed in cooling and dry is medium | What you notice when the reading is wrongAirflow weakens while you are still in the room, and the unit reads as underperforming |
| Response once the room reads as emptyAirflow steered away from people | Where the makers document itDaikin: louvers move so airflow is not directed at a detected area. Mitsubishi Electric: Indirect, set per air outlet on a cassette | What you notice when the reading is wrongCold air lands on you anyway, or leaves the seat you are actually using |
| Response once the room reads as emptyAirflow aimed at people | Where the makers document itMitsubishi Electric: Direct, plus an Even setting that learns where an occupant spends most of the time | What you notice when the reading is wrongThe draft follows a chair nobody is sitting in |
| Response once the room reads as emptyUnit stopped | Where the makers document itMitsubishi Electric: no-occupancy auto-off, set between 60 and 180 minutes in 10-minute steps | What you notice when the reading is wrongCooling ends while the remote display still indicates the unit is in operation |
The bedroom is where the assumption breaks
A bedroom gives this feature the least to work with. Stillness is the point of the room, and stillness is the condition these manuals list as a reason a person goes undetected. The machine is asked to separate an empty bed from an occupied one using the signal a sleeping body is least able to produce. Every other room in a flat is an easier problem than this one.
Bedding removes another margin. Mitsubishi Electric lists a person wearing heavy clothes whose skin is not exposed among the cases its sensor may miss, and a blanket presents the sensor with the same problem as heavy clothing. The same manual names positions the sensor cannot cover: along the wall on which the air conditioner is installed, directly under it, and anywhere an obstacle such as furniture stands between the person and the unit. Bedrooms in flats here put the bed in one of the first two more often than not, and a wardrobe supplies the third.
On some units the feature also takes something away. Daikin's Singapore manual for one wall-mounted range states that night set mode will not go on during use of intelligent eye operation. A household already running the night setting is choosing between two features, not adding one. The manual is the only place that says so, and nothing printed on the remote itself suggests a conflict exists.
This is where the feature arrives as a complaint instead of a benefit. Cooling that fades in the small hours, or a unit found off in the morning while the remote still shows it running, both have an occupancy explanation to rule out before anyone talks about gas. Mitsubishi Electric states the second case outright: even when the unit has stopped on its no-occupancy timer, the display of the remote controller remains to indicate the unit is in operation. Anyone weighing sleep mode vs auto mode for the night has a third setting to account for.
What decides whether it earns its place
The saving belongs to your household, and not to the machine. It comes entirely from the share of running hours in which the room is genuinely empty. A family that switches the aircon on when somebody walks in and off when they leave has already collected that saving by hand, and a sensor has nothing left to recover from them. The feature is worth most to the households least disciplined about switching off, which is an uncomfortable thing for a spec sheet to admit.
The households where the share is large are specific ones. A living room unit left running through an afternoon while people move in and out. A study that empties for errands and fills again after dinner. A unit switched on before anybody is home. The households where the share is small are just as specific, and a bedroom unit that runs from bedtime until morning is the clearest of them. Nobody leaves that room empty for an hour while the machine runs.
The failure also runs in the other direction, and Daikin prints the list. Its manual states that the sensor also mistakenly detects pets, sunlight, fluttering curtains and light reflected off of mirrors as passersby. Fujitsu General publishes a matching list, and adds an oscillating fan to it. A room with a cat, an afternoon sun patch, or a curtain that lifts under the airflow may never register an absence at all. In that room the feature costs nothing and does nothing, which beats the opposite outcome and is still not what the brochure described.
No saving percentage appears on this page because none of these documents carries one. The manuals state what the machine will do, and stop there. The energy label will not settle it either. NEA sets the tick rating from a weighted COP, built from full-load and half-load measurements in a laboratory. A feature whose value depends on when your rooms empty falls outside what that test can report. What any of it takes off a bill depends on hours only your household can count, and on how often the sensor reads the room correctly. A percentage quoted without both of those is describing somebody else's flat.
- The model number, not the brand name. Which instrument is fitted varies from one model line to the next, and changes again between production years.
- What the manual says the sensor detects. Movement and floor temperature are different claims, and the feature name rarely settles which of the two is fitted.
- What the unit does once it decides the room is empty, and after how long. A setpoint shift and an automatic shutdown are different things to live with.
- Whether it can be switched off, and whether turning it on blocks a setting the household already uses.
- Where the sensor sits on the indoor unit, and what stands between it and the seats people actually use.
- Whether that room is ever occupied and then empty inside the same running session. If it never is, the feature has no work to do.
Ready to get started?
Tell us what’s going on. Symptoms, setup, photos, anything we should know. We’ll assess and come back with the right next step.