Here a link from the Swedish AES tests. The main link with more detail is gone and was not archived. This one has the basics of how they did the testing.
Here is a simple diagram which should make it clear what they were doing.
They used a complex load modeled after a difficult loudspeaker. You can use a power resistor for a first run if you wish. If you don't understand this ask questions prior to doing such a test.
You feed the input of the DUT (device under test or in this case an amplifier) a signal from the preamp. You load the output with a power resistor. You tap across that load with a voltage divider to reduce the voltage. You reduce it equal to the gain of the amplifier. If the amp has 26 db of gain then that is a ratio of 20:1. So once you have done this you can feed the power amp a signal and out comes a signal from the other end equal to the signal level you input. Unity gain.
So you take this unity gain signal and send it onto the power amp you have powering your speakers. With what they have labeled volume 2 above and the switch prior to it, that one switch can switch the DUT in and out of the circuit. A perfect amp would sound the same as wire. Now there are plenty of little details of arranging this you'll have to work out. At one time back in the 1990's I had things setup so I could put an amp in and try this in a couple minutes. After that I could switch or have someone switch the amp in and out of circuit. I've not done this since around the year 2000 or so. I don't know if switching amps will cause any issue with the switching frequency noise input into another amp. I do know that unlike the diagram most switching amps are differential instead of one speaker terminal being ground. You'll have to take that into account with how you wire all of this up.
So if someone had this all setup it is no big deal to use, but there is a bit of bother getting that ready to go. Also, I only found one amp that was pure straightwire with gain doing things this way. That was a Spectral DMA-50. Lots of SS amps of quality came pretty close, but only the Spectral was so good you simply couldn't tell if it was in circuit or not. I don't know how many amps the Swedish AES publication tested this way. Or when they stopped doing this. They only found one amp passed their sighted testing and it failed the blind portion. It was a large Audio Research solid state amp. They found one amp that passed both sighted and blind testing for pure transparency and that was a Bryston SST.