How do you tell if EQ isn't "clean"?

[Unaspected]

Seems weird if a regular EQ was using FFT. Maybe something like soothe or gulfoss uses FFT (no clue), if you want to call those EQs?

Is it? If you want to make a dip of -6dB at 1kHz, you somehow have to "extract" from the complete audio waveform within the time window the EQ uses the part that creates the sound around 1kHz from everything else. Going by Wikipedia "... Fourier analysis converts a signal from its original domain (often time or space) to a representation in the frequency domain and vice versa."

But I'm completely clueless about the details, so I might be wrong.

You can easily design a FIR filter using IFFT or IDFT - [that is, the inverse of the Discrete Fourier Transform]: Where the spectrum can be directly edited to adjust the magnitude of the frequency bands - called "bins". There are quite a few aspects to take into design and this method will not [likely] be as precise as pertaining to attenuation of a specific frequency as the more we divide the frequency spectrum, the more latency we will incur. Whilst, the fewer bins, the less we know what is actually happening between each bin.

Then there is the matter of how strange it can sound to attenuate just a single sinusoid. We've all heard how weird a low bit rate mp3 file can sound in high frequencies when very low "masked" frequencies are completely stripped. So we might consider designing our FIR filter around known analogue types to create maybe a peaking filter or band reject filter.

[Unaspected]

As for adjectives like "clean" or "musical," I think it's mostly pretentious nonsense. An EQ has no degree of "cleanliness." Or godliness.

Active EQs generate noise and are considered to be less clean than passive EQs. It's unlikely this aspect would be modelled in a plugin so the only qualifier of clean or otherwise would be potential to saturate.

[applefrozenyogurt]

I have never been able to tell the difference between two EQs...