There's different levels of sophistication you could give to this question, but at the most basic level: empirically, general relativity is an extremely successful theory of gravity and agrees with observations every time we can test it. So, we tend to trust it's predictions as a reasonable, default hypothesis even when we cannot directly verify them.
General relativity predicts gravitational waves have a spin-2 character. While we've never directly measured the polarization (beyond ruling out some extreme, strawman possibilities), many other predictions of GR with respect to gravitational waves have turned out to be correct (such as the spin-down rate of the Hulse-Taylor pulsar, the direct detection of gravitational waves by LIGO/Virgo from compact binary systems, the speed of gravitational waves, etc). According to the normal rules of quantization, classical gravitational waves become gravitons. So we expect spin-2 polarized classical gravitational waves to correspond to a spin-2 graviton after quantization.