It potentially or ideally elevates consensus and moderate views, but since AIs are privately and generally corporately owned, and corporations are mindless monsters controlled by psychopathic shitweasels, the reality is that it will be (and to some notable degree already is) bent to the promotion of whatever toxic agenda the shitweasel owner(s) prefer.
And in point of fact, it's made just that much more dangerous specifically because people tend to assume it represents some sort of "expert consensus and moderate views."
We're actually better off with populist and polarizing, since at least among that clamor of voices, the truth can still be shared. Under corporate AI, any truths the shitweasels want hidden will vanish entirely and any lies they prefer will be front and center.