Depends on what the generator is calibrated in!
Certainly in the UK, and I believe a lot of Europe, historically generators were calibrated in terms of the EMF i.e. the open circuit volts. The US, on the other hand, calibrated in terms of the actual volts into the correct load - which is half the EMF.
Then the radar/ECM people came into the game. For them, it's actual power that matters and so we found generators calibrated in dBm, which is the power in 50 ohms that it would be if the load was matched. In most cases at HF, the receivers have quite a high input SWR, but provided the cables are short at HF, the results are repeatable enough.
Marine radio standards required on AM that switching off 30% modulation gave a 10dB change in output, while the specifications were all for EMF voltages: below 3.8 MHz, the dummy antennas were specified - for 1.6 to 3.8 MHz, it was 250pF in series with 10 ohms, so it was an interesting and lossy little network wthat presented 50 ohms to the generator and source impedance of 10 ohms and 250pF in series to the rx. I can't remember the lower frequency dummy antennas though - it may have been 6 ohms and 350pF at 500kHz.
So with multiplicity of methods, it has to be defined which is used!
Incidentally, there's a good article in the latest QEX on antenna noise, although his reference to the ITU-R Rec. P372 -7 is several years out of date: the current version is P.327-10
That is the beauty of standards, there are so many to choose from...
My generators are in DBm and microvolts into 50 ohms.
I also use a 6 DB pad for 'hard' microvolts.
When I do a sensitivity check, I switch 30% modulation of 1000 Hz off for a 6 DB drop in the audio output level.
Most of the receivers I check are 1 microvolt for 6 DB drop.