How to Test a Home's Alarms Correctly

Why this matters

A tech who presses one test button, hears a horn and writes "alarms OK" has tested exactly one horn. The house can still have a bedroom unit past its life, a basement unit that never joined the interconnect, and a hallway unit whose chamber is sealed shut with paint overspray, and every one of those passes a quick press. When a shop offers alarm testing as part of an end-of-job check or a maintenance visit, the customer reads "tested" as "will wake us up." This card is the procedure that earns that word, and it is also honest about the part of the job a test button cannot do.

The stance behind it is owned elsewhere: see related, Why Every Trade Owes the Life-Safety Devices in a House. This card is the method.

What the test button proves, and what it does not

Get this straight before you touch anything, because it decides what you can write on the ticket.

On a smoke alarm, the test button runs the alarm's own electronics and sounds the horn. On many models it also checks the sensing circuit in some simulated way. It does not prove that smoke can get into the chamber. A unit whose vents are packed with drywall dust, painted over or covered with a forgotten bag will sound on the button and stay silent in a real fire.

On a carbon monoxide (CO) alarm, the test button proves the horn and the circuit, and nothing about the sensor. The sensor is a small electrochemical cell that wears out over years, and the button cannot tell you whether it still responds to gas. A field check of the sensor needs the manufacturer's listed test gas and procedure, which is a heating tech's or alarm contractor's job with the right kit, not something improvised.

On an interconnected system, pressing one button proves that unit can send a signal and that the units you heard can receive it. It does not prove the others can send. Proof that every unit can send means pressing each of the others in turn.

So a complete test has four parts: every horn, every unit pressed as an interconnect initiator, every manufacture date, and the chamber where the manufacturer allows a smoke aerosol. Anything you could not prove, you write down as unproven.

The steps

1. Tell everyone first, including anyone who monitors the house. Tell the occupants which rooms will sound and for how long, and ask about anyone sleeping, a baby, a dog that bolts, or someone with a hearing aid or a heart condition. Then ask one question that saves a fire truck: "Do any of these report to an alarm company?" Some homes have system smoke detectors wired to a security panel, and a test there can send a fire signal to a central station that dispatches the fire department. If the answer is yes or unsure, the customer calls the monitoring company and has the account placed on test before you start, and takes it off test when you finish. Skip this and you either scare the household or roll an engine to a house with no fire.

2. Inventory every device before you press anything. Walk the house and list each unit by location, sensing function (smoke, CO, combination, heat), and power (hardwired with battery backup, sealed battery, replaceable battery, plug-in). The inventory is what turns "the alarms work" into a count you can check: if you found seven devices and heard six, you know which one to chase. Skipping it is how a basement or garage unit goes untested for years. Use a stepladder with all four feet on the floor and someone steadying it for ceiling units; reaching over from a chair is how this job hurts people.

3. Press and hold each test button until the horn sounds, then release. Most units sound within a few seconds; some need a longer hold, and the manual says how long. Use a broom handle on high ceilings rather than climbing higher. Wear ear protection or step back and cover your ear: a residential smoke alarm is built to produce at least 85 dBA at 10 feet under UL 217, and you are inches away. A unit that does not sound on a held press has failed outright. Do not swap the battery and call it fixed until you know why it was dead, because a unit with a dead backup on a hardwired circuit may also have lost its AC feed.

4. Prove the interconnect from every unit. Press one unit (the initiator) and have a second person, or a walk, confirm that every other interconnected unit sounds. Then press each of the others in turn and confirm the rest follow every time; a unit whose receiver works and whose sender does not only shows up when it is the one pressed. Interconnect may be a wired third conductor or a wireless link, and both fail silently: a unit replaced with a standalone model, a wireless unit that dropped its pairing, a cut wire behind an addition. A unit that sounds on its own button but not when the initiator is pressed is not on the interconnect, and it is a reportable finding even though the unit itself works.

5. Read the manufacture date on every unit. It is printed on the back or side of the unit, so twist it off its base to read it. On a hardwired unit, support it rather than letting it hang on its connector; if the connector has to come apart to read the label, the unit runs on battery until you plug it back and reseat it, which you do the moment you have the date. NFPA 72, in the edition your jurisdiction has adopted, calls for smoke alarms to be replaced no later than 10 years from the date of manufacture, and the manufacturers say the same in their manuals. CO alarms carry their own life, usually printed as a replace-by date on the label; follow the label, not the smoke-alarm rule. An undated unit is treated as past its life. Date reading belongs to its own card: see related, Reading the Manufacture Date and the Ten-Year Replacement Rule.

6. Test the smoke chamber with a listed aerosol, only where that unit's manual permits it. Some manufacturers approve a listed smoke-alarm test aerosol, sprayed from the distance and for the duration the can and the alarm manual state. Some say not to use any aerosol, because the residue can contaminate the chamber. Read the manual or the label first and record which units you could not chamber-test. Never use an open flame, a match, a candle or a smoldering rag: it is a fire source on a ladder under a ceiling, it deposits soot that can ruin the sensor, and it proves nothing the aerosol would not. Never use car exhaust or any combustion source to test a CO alarm.

7. Reseat, re-press, and leave every unit armed. After any unit came off its base, reseat it until it locks, confirm the AC power light on a hardwired unit, and press its button once more. The return-to-service check is the part that gets skipped: a unit left hanging by its connector, or a base twisted half a turn short, looks installed and is not.

8. Record the results device by device, and hand the customer the findings in writing. One line per device, with the four results and anything unproven. The record is what protects the customer, and it protects the shop when someone later asks what was checked.

Worked example: a two-story house with a basement

A plumber finishing a water heater swap offers the end-of-job alarm check. The occupants are home, the dog is shut in the yard, and the customer confirms nothing is monitored. The inventory finds eight devices: six smoke alarms, all expected to be hardwired, in the three bedrooms, the upstairs hall, the main-floor living room and the basement, and two plug-in CO alarms, one in the upstairs hall and one in the basement near the furnace and the new water heater.

ALARM TEST RECORD - 2 story + bsmt - 2026
LOC        TYPE   POWER   MFG   BTN   INTRCN  AEROSOL
Bedroom 1  smoke  HW+bat  2021  ok    ok      ok
Bedroom 2  smoke  HW+bat  2015  ok    ok      ok
Bedroom 3  smoke  HW+bat  2021  ok    ok      ok
Up hall    smoke  HW+bat  2021  ok    ok      NO RESP
Living     smoke  HW+bat  2021  ok    ok      ok
Basement   smoke  bat     2023  ok    NO      not perm
Up hall    CO     plug    2020  horn  n/a     n/a
Basement   CO     plug    2017  horn  n/a     n/a
Each unit pressed as initiator. Basement: only itself.

Reading it line by line:

• Pressing Bedroom 1 sounded five of the six smoke alarms. The basement stayed silent. Pressing each of the other four upstairs and main-floor units gave the same five. Pressing the basement unit sounded only the basement. So the link fails in both directions, and the reason is visible once the unit is off its ceiling: it is a battery-only alarm with no connector, put up during a basement finish. It works, and a basement fire would not sound a single bedroom unit two floors up.
• Bedroom 2 was made in 2015, which in 2026 is 11 years, past the 10-year line. It passed every test. It still goes on the replace list, because the date rule exists for sensors that degrade before they fail outright.
• The upstairs hall smoke unit sounded on its button and on the interconnect, then gave no alarm to the listed aerosol within the time its manual gives. Its vents carry a skin of ceiling paint from a recent repaint. This is the case the whole procedure exists for: a unit that would have passed a button-only test.
• The basement smoke unit's manual does not approve aerosol testing, so it was not chamber-tested, and the record says so.
• Both CO alarms sounded on the button. The record says "horn" rather than "ok" because that is all a button proves. The basement unit's label reads replace by 2027, so it is in service and noted for next year.

The written findings to the customer: replace the Bedroom 2 alarm (past 10 years) and the upstairs hall alarm (no response to smoke); the basement smoke alarm is not interconnected with the rest of the house; both CO alarms sounded on test, their sensors cannot be proven by a button, and the basement unit reaches its replace-by date in 2027. Three of the six smoke alarms need action and the plumber has replaced none of them, because on hardwired interconnected units the replacement and the basement interconnect are an electrician's work where the state or locality licenses it. The plumber's job was to find it and say so.

What would change the call. If the basement unit had been a wireless model of the same family, the fix might be re-pairing it per its manual rather than new wiring, which some shops' handymen can do. If the house had been monitored and nobody asked, the first press would have sent a signal to a central station. If the hall unit's manual had forbidden aerosols, its paint skin would have gone unproven, and the honest entry would be "chamber not tested, vents visibly painted, recommend replacement" rather than a pass.

How to know you did it right

• The device count in your record matches the count you walked, and every row has a result in every column, including "not permitted" or "n/a" where that is the truth.
• The interconnect column shows every unit pressed as an initiator, not one or two.
• Every manufacture date is a year, not "looked fine."
• No CO alarm is recorded as "sensor OK" on the strength of its button.
• Every unit you took down is back on its base with its power light on, and pressed once after reseating.
• If the house is monitored, the customer has confirmed with the monitoring company that it is off test.

References

• NFPA 72, National Fire Alarm and Signaling Code, in the edition adopted by your jurisdiction (smoke alarm replacement at 10 years from manufacture)
• UL 217, the listing standard for smoke alarms (sound output)
• Manufacturer installation and use manuals for each alarm (hold time, aerosol permission, CO replace-by date)
• See related: How to Find the Chirping Alarm and Tell What the Chirp Means
• See related: Smoke Detector Placement + Types Reference