I agree with this to some extent, how direct should the harm be before you stop distributing it?

I don't think there is an easy, obvious line there—direct calls to harm (murder these people) shouldn't be accepted, and I believe even beyond that there should be limits, but find that line and enforcing that is very hard.

You are downvoted but this is exactly what happening in the Arabic sphere of discussion and nobody really cares because Google, like most of the discussion here, is very America centric. This is not even about truth but rather about calls for murder which are against the law in many countries.

In some way though, I am happy that this "diversity of thought" exists because that's the only thing that can stop companies like Google from becoming the arbiter of truth. It is just sad that a lot of those countries are also ok with calls for murder and violence.

That is a rather curious argument. Google's management come from Silicon Valley and statistically will have a left bias. It is quite likely that a substantial number of them genuinely believe that, say, attending a right-wing rally is similar to presenting Mein Kampf as truth with no evidence. If they start censoring one it is pretty likely they will start censoring the other.

Also, if they use controversy as their metric instead, they're going to start squashing minority views. It is actually much harder than you might think to decide what is and isn't obviously true, particularly when there is a team of moderators involved (or team of people calibrating the robots which is maybe how Google does it).

Censorship is really dangerous.