Grounded circuit (green/bare ground wire wired properly):
- Push "Test" on the GFCI protective device, and it trips -> PASS
- Push "Test" on the GFCI protective device, and it does not trip -> FAIL
- Push "Test" on a plug-in GFCI tester, and it trips -> PASS
- Push "Test" on a plug-in GFCI tester, and it does not trip -> FAIL
A fail here indicates the GFCI unit is probably defective.
Ungrounded circuit (green/bare absent or defective):
- Push "Test" on the GFCI protective device, and it trips -> PASS
- Push "Test" on the GFCI protective device, and it does not trip -> FAIL
- Push "Test" on a plug-in GFCI tester, and it trips -> FAIL FAIL FAIL
- Push "Test" on a plug-in GFCI tester at the GFCI receptacle, and it does not trip -> PASS
- Push "Test" on a plug-in GFCI tester at a downstream protected outlet, and it does not trip -> MEANINGLESS
Fail at the GFCI device probably indicates it is defective.
Fail on the plug-in GFCI tester (i.e. it trips!) indicates they have bootlegged ground at the GFCI receptacle- attached the neutral wire to both neutral and ground. It will seem to work at the GFCI, but is still dangerous.
Bootlegging ground at a downstream GFCI receptacle is a mistake, because one of several electrical faults could put 120V on the the grounds, e.g. the cover plate screws or a machine chassis. However this is difficult to detect, since a properly wired downstream receptacle will behave exactly the same way. This means for ungrounded downstream receptacles, plug-in tester testing is completely meaningless.
Once you have settled the question of bootlegged grounds, here's how you test an ungrounded GFCI. Plug your GFCI tester into your handy dandy 2-3 prong "cheater" - the kind with a short green wire as a pigtail. Extend that green wire all the way to a reliable ground source, e.g. the panel in the basement. Now, the GFCI tester should work normally, since you have rigged a proper ground to it.
GFCI protection is pretty effective, and I would be confident in an ungrounded circuit if it has GFCI protection. However if you are unable to get the external device to trip, you'll need to pop the cover off and see if the ground is present, missing or bootlegged.
If you don't trust a ground, don't connect to it - use GFCI instead.
If a GFCI won't trip with its own internal test button, it is duff. Into the trash it goes.
Here is how GFCIs and ground wires are supposed to relate to each other -- or to be more precise, how they are not.
To the left you see unprotected hot and neutral coming in, out the right you see protected hot and neutral, which I've recolored. Ground sails right by unconnected (normally). Obviously, if your ground is corrupt/defective, this is bad news indeed.
"Wait, all the GFCI's I've ever seen have a ground screw." No. That thing you call a GFCI is actually a GFCI+receptacle combination device. It provides a GFCI module, and also two sockets (wired past the GFCI). The GFCI can't use ground. Look at a GFCI breaker, it doesn't even have access to ground. The ground is for the sockets. This means effectively, that ground screw is on the "protected" side of the GFCI.
So if the ground is bad, where should you cut the ground? Before it reaches any protected loads, and remember, the ground screw on a GFCI+receptacle combo serves the protected loads.
Should you do anything creative like tie the protected-side ground into protected-side neutral? No No No! This post of mine explains how that utterly defeats the GFCI protection. Wrap the ground wires with tape so they can't touch, and don't use them.
Detecting tied ground-neutral
If your panel is set up this way, the single easiest way to test your neutral-ground isolation is to disconnect your neutral-ground bond in your main service panel. Now the only thing connecting neutral and ground is that long path of dirt between your grounding electrode system and the pole transformer's. If you also unhook your grounding electrode, your house's internal grounds should be fully isolated from neutral, and should megger out at a couple megaohms. (Don't megger things in residences though, it could fry electronics).
Or test circuit by circuit. It's a simple thing, on any given circuit there should be 0.000 amps of current flow on the ground wire. Nothing is supposed to use ground but test equipment. Now, if neutral and ground are tied together, current follows all paths in proportion to their conductance (1/resistance) so a significant fraction of current will take ground instead of neutral (assuming there is a load). Obviously a GFCI will detect the shortfall, but a clamp meter will detect the ground current directly.
I don't agree with that video's claim of nearly 1 ohm between neutral and ground. Copper wires have much better conductivity than that unless he has many hundreds of feet of wire between his lab and his main panel. There may be something peculiar going on in his test lab, or he is misunderstanding or misusing the equipment. You shouldn't have 1 ohm neutral-ground, that would limit dead-short current to 120A, which would not flow enough current to safely magnetic-trip a breaker.
In any case, a clamp ammeter around the N-G bond (or a circuit's ground wire) would soon show if any AC current was flowing.
Best Answer
It is normal for a plug-in GFCI tester to not work, if the receptacle is not grounded. GFCI testers are not magic.
GFCIs work by comparing current on the "hot" wire to current returning on the "neutral" wire. If all is well, they should be equal. if they are not equal, some current is traveling down a third path possibly through a human.
If a third path does not exist, then tripping a GFCI is impossible. An external tester cannot trip a GFCI if it only has access to hot and neutral. The GFCI itself can test itself by using an internal path which goes around the LINE and LOAD terminals, both of which it has access to. An external tester can't do that.
So failure to trip at a downstream outlet does not tell you anything one way or the other about whether ground has been bootlegged there. It only tells you ground is not present. Such a receptacle should have "GFCI protected" and "No Equipment Ground" stickers and will need those to pass home inspection.
To use an external tester at such a location, get a common 3 to 2 prong cheater, the kind that has a 3 inch green pigtail meant to be attached to the receptacle screw. Extend that using whatever wire you have on hand, all the way to the panel, attach it to the building's grounding electrode system. That becomes the third path.
You are allowed to retrofit proper grounds to a receptacle without disturbing the hot/neutral wiring, and you don't even need to use the same route. (Also remember metal conduit is a ground path). However if a circuit has GFCI protection upstream, I would call such an upgrade a low priority. In some ways, GFCI protects better than actual grounding.