Smoke alarm replacement invoices: the date on the back of the alarm

A vendor bills for “replaced smoke alarms, building C, 36 units”, with a set of photos attached: white alarms on white ceilings, one after another. The work may be exactly as billed. But a photo of a new alarm on a ceiling looks the same in every unit, on every job, in every year. The alarm itself carries a better record, printed on its back: the date it was made. This piece is about what the ceiling photos can show, what that date adds, and a few checks that take about five minutes. It is not legal advice.

Why the date matters

The U.S. Fire Administration says smoke alarms “need to be replaced 10 years from the manufacture date,” and for 9-volt and hardwired alarms, “Replace the entire smoke alarm every 10 years.” Sealed 10-year lithium alarms are to be replaced “according to the manufacturer’s instructions.” The clock runs from when the alarm was made, not from when it was bought or installed. And the date is easy to find: a city fire-safety page republishing the USFA’s material puts it simply, “Look at the back of the alarm for the date of manufacture.”

So each alarm on an invoice has two dates worth seeing: the one on the alarm that came down, and the one on the alarm that went up.

What the ceiling photos can show

• That an alarm is mounted there. A photo shows an alarm on a ceiling or wall, and roughly what kind it is.
• Sometimes which unit. If the frame includes something only that unit has, like the door number, a light fixture or the view from a window, the photo can be tied to a place.
• Sometimes when. If the files still carry the camera’s own clock, the dates should fall in the window the vendor says they worked, and on or before the invoice date.

What they can’t

• That the alarm is new. A ten-year-old alarm and one made last month look alike from below.
• That the old one needed replacing. The ceiling photo says nothing about the age of whatever was there before.
• How many were replaced. Thirty-six near-identical ceiling photos can be thirty-six alarms, or a few alarms photographed thirty-six times, or photos from another building.
• That it works. A photo isn’t a test. The USFA says to test alarms monthly; a test is something someone has to do, not something a picture records.

What a photo of the back adds

• The old alarm’s age. A manufacture date more than ten years back supports the replacement. A date two years back is a question: why was a young alarm replaced?
• The new alarm’s age. The date should be recent. A new alarm with an old date is worth asking about. Stock can sit on a shelf, so it is a question, not a finding.
• The model. The label shows what was actually installed: battery, hardwired, or a sealed 10-year unit. If the invoice priced one type, the label shows which type went up.
• A count you can check. One back-label photo per alarm billed, each with a date, is harder to pad than a stack of ceiling shots.

Be honest about what this can do, too. A photo of a label is only a photo. It shows that someone photographed the back of an alarm with that date. It doesn’t show which unit the alarm came out of or went into. Pair it with a photo that includes something unique to the unit, and keep the old alarms until the invoice is paid if you can.

Five minutes of checks

1. Put it in your vendor terms: for every alarm replaced, a photo of the back of the old one and the back of the new one, plus one photo showing the alarm in place in the unit.
2. Check the count. The number of new-alarm label photos should match the number billed. If the vendor supplied the alarms, the supply house receipt should cover that quantity too.
3. Read the dates. Old alarms past ten years support the work. Old alarms that were nowhere near ten years old need an explanation, such as a failed unit, damage or a code change.
4. Ask for the original photo files, not pictures pasted into an invoice PDF, which have usually lost their camera record. Why, and how to ask for the originals.
5. Drop this job’s photos in with the vendor’s earlier ones. The checker marks any two files in the same batch that are the same file, even under different names. One label photo used for three units, or a photo reused from last year’s job, shows up that way. A second photo of the same alarm won’t, because it is a different file, so read the dates and models too.
6. Write the dates into your own unit records. Then the next inspection, and the next vendor, starts from your record instead of theirs. Where to keep inspection photos so they stay useful.

Whether someone used AI on it

An AI picture of a smoke alarm on a ceiling is easy to make now, and some generators write a mark into the file saying so. The checker reads those marks, and the notes some phone editors leave when generative AI was used to alter a photo. A file with no such mark is not proven real. Most fakes and most genuine photos look the same to a file reader once they have been through a chat app or a screenshot. The dates printed on the alarms, and the old alarms themselves, are checks that don’t depend on the photo at all.

The limits, stated plainly

• A ceiling photo shows an alarm in place. It doesn’t show the alarm’s age, what was there before, or that it works.
• A photo of a manufacture date is corroboration, not proof. It doesn’t show which unit the alarm belongs to, and new stock can carry an older date.
• The ten-year guidance quoted here is the U.S. Fire Administration’s, read in October 2026. We have not covered which alarms your state or city requires in a rental, where they must go, or who must replace them. Check your local code.
• A photo with no camera record is not a fake. Text messages, chat apps and many portals remove the record from genuine photos every day.
• Our checker gives no confidence score and does not judge a picture by looking at it. It reads what is in the file, in your browser, and uploads nothing. A file with no AI marker is not proven real.
• This is not legal advice.

Related: water heater invoices and the data plate, and what a contractor’s “after” picture actually proves.

---

ImposterShield reads the evidence inside image and video files. It runs in your browser and uploads nothing. If a file has been stripped, it says so. If it has not read a file, it says that too.