You flip the switch, the room goes dark, and a minute later you notice it: a dim, stubborn ember where the LED used to be. Not bright enough to read by. Just bright enough to bother you. You swap the bulb. Same thing. You start to wonder if the wiring is bad, or worse, if something upstream is quietly leaking energy into a circuit that is supposed to be dead.
The good news is that nothing is broken. The bad news, for your peace of mind, is that the light was always leaking a tiny current — you just never had a bulb sensitive enough to show it. The incandescent bulb that lived in that socket for forty years swallowed the same trickle without a flicker. The LED can't. Understanding why is a small lesson in AC physics that turns a mysterious callback into a two-minute fix.
The bulb that used to hide the evidence
An incandescent bulb is a coil of tungsten with a resistance of maybe ten to a few hundred ohms when cold. It is, electrically, a straightforward resistor. To make it glow even faintly, you have to push real current through it — tens of milliamps at least. Anything smaller than that dissipates as an imperceptible warming of the filament and vanishes.
An LED bulb is a completely different animal. Behind the diffuser is a small switch-mode driver: a bridge rectifier, a smoothing capacitor, and a controller that turns raw 120-volt AC into the low-voltage DC the diodes actually want. That driver is a high-impedance front end with a capacitor sitting right at its input, waiting to be charged. It does not need milliamps to react. A few microamps, dribbling in over a few seconds, is enough to charge that input capacitor up to the point where the LEDs briefly conduct — and you get a glow, or a slow, rhythmic wink every several seconds as the capacitor fills and discharges.
So the question isn't really "why does the LED glow." It's "where are those few microamps coming from, on a wire that the switch just opened?"
Your walls are full of accidental capacitors
Here is the part that surprises even seasoned electricians the first time they think it through. Two insulated conductors running side by side through a wall — the switched hot and the neutral, say, sharing the same cable for a twenty- or thirty-foot run — form a capacitor. Not a good one. But a real one. Two conductors separated by insulation is the literal definition of a capacitor, and the longer they run in parallel, the more capacitance they build up. A typical run might land somewhere on the order of a hundred picofarads.
Capacitors pass AC. That's their whole personality. Even with the switch open and no metallic path for current to flow, the alternating voltage on the live conductor couples across that tiny capacitance to the switched leg on the other side. The opposition a capacitor offers to AC is its reactance, and the formula is worth carrying in your head:
Xc = 1 / (2πfC)
Run the numbers for 100 picofarads at 60 hertz and you get a reactance of roughly 26 million ohms. Across 120 volts, that permits a leakage current of about four or five microamps. To an incandescent filament, four microamps is nothing — it doesn't even register as heat. To the input capacitor of an LED driver, four microamps delivered patiently is a slow drip filling a bucket. Eventually the bucket tips, the diodes flash, and the cycle repeats. That is capacitive coupling, and it is the single most common reason a switched-off LED glows.
The other usual suspect: the switch that lights up
The second big cause is a switch that was chosen to be helpful. An illuminated or "locator" switch has a tiny neon lamp or LED inside the toggle so you can find it in the dark. That indicator needs a sip of current to stay lit, and in most designs that current flows through the load — through your ceiling fixture — even when the switch is off. With an incandescent bulb in the socket, the sip was invisible. With an LED bulb, that same milliamp-or-less trickle is more than enough to keep it faintly awake.
A close cousin of this is the digital timer, motion sensor, or smart switch that has no neutral of its own and instead "steals" a small keep-alive current through the load to power its own electronics. Same physics, same result: a path for a little current to reach a bulb that is supposed to be dark.
And there is a third, less benign possibility worth naming: a miswired circuit where the switch interrupts the neutral instead of the hot. In that arrangement the fixture stays energized with respect to ground even when "off," and the leakage has an easy route home. That one is not just a nuisance — it's a shock hazard for whoever changes the bulb, because the lampholder is live with the switch off. NEC 404.2(A) is specific that switches shall break the ungrounded (hot) conductor for exactly this reason. If you find a switched neutral, you've found something to correct, not tolerate.
How to make the ghost go away
Once you know the mechanism, the cures are obvious and they all do the same thing: give those stray microamps somewhere harmless to go, or cut them off at the source.
The cleanest fix for capacitive coupling and illuminated-switch leakage is a bleeder — a small resistive or capacitive load wired across the fixture, in parallel with the LED. Manufacturers sell them as "LED anti-flicker" modules or "CFL/LED capacitors." The device presents a low-enough impedance that the few microamps of leakage flow through it instead of accumulating in the LED driver, and the driver never reaches its turn-on threshold. The glow stops. The trade-off is a hair of wasted energy, which is negligible.
If the offender is an illuminated or no-neutral switch, the honest fix is to replace it with a standard switch, or to run a neutral to it so its electronics stop borrowing current through the lamp. If it's a switched neutral, rewire it so the hot is switched — full stop.
And sometimes the simplest move is to change the bulb brand. LED drivers vary enormously in how much input capacitance they carry and how twitchy they are at low current. A different make with a slightly leakier internal bleeder of its own may simply refuse to glow on the same circuit that lit up its predecessor.
The quiet lesson in a glowing bulb
What makes this problem satisfying is that it isn't a fault at all. It's your wiring behaving exactly as physics says it must, revealed by a load finally sensitive enough to see it. The current was always there. Ohm's law, capacitive reactance, and the difference between a resistive filament and a switch-mode driver were always in play. The LED just turned up the contrast on a phenomenon the incandescent bulb spent decades papering over. Diagnose it well and you're not chasing a defect — you're reading the electrical shadow the building casts.
That's the kind of call where the answer lives in the numbers, not the guesswork: what's the reactance of a hundred picofarads, how much leakage does that permit, is that enough to wake a driver. Voltly keeps those calculations — reactance, voltage drop, ampacity, conduit fill, box fill — and the NEC references behind them in your pocket, working fully offline in a basement or a mechanical room with no signal. When the mystery on the job is really just a formula you haven't run yet, it's faster to reach for the answer than to argue with the ghost.
If you'd rather carry the reference than the drawer of dog-eared tables, take a look at Voltly.