I find trusting AI to detect misinformation to present essentially the same problem as using AI to spread it - both cases require implicit trust in an untrustworthy system, and removes the impetus to educate people and have them practice critical thinking and skepticism (as opposed to cynicism and contrarianism) themselves.

The most effective system I've seen so far, ironically (given who now runs the site) is the community notes on Twitter. But even that gets gamed by activists and bad faith actors.