1/r² is for an isotropic radiator, but Voyager 1 mounts (thankfully) a high gain directional antenna. The gains seem to be 36dBi or 48dBi, depending on the frequency [1]. This means that when pointed towards Earth, the signal power is actually between 4000 to 63000 times stronger than what that formula would give you. It's still tiny, but not as bad as that would suggest.

[1]: https://space.stackexchange.com/questions/65801/what-are-the...

No, it is for any radiator. The gain of the antennas has absolutely nothing to do with the free path loss, it just changes the effective radiated power or the concentration effect of the antenna because of antenna related gains. Every kind of antenna has its own specific gain and corresponding losses diagram based on which way the source is relative to the antenna, with the baseline being an omnidirectional radiator (which no antenna really is, but we assume it is).

GP was very specific when they mentioned free path loss: the implication is that if you know what free path loss is then you also know that those losses are accurately modeled with that formula, no matter how large your ERP or receiver antenna gain is. At those distances it would make zero sense to use an omnidirectional antenna you will need every bit of gain you can get.

For clarity, it’s still proportional to 1/r², but with a much larger multiplier than one.