Yes, trying to solve this problem by creating a definition of "misinformation" sufficiently precise to act on it for all possible articles, then trying to remove it somehow is probably an AI-complete problem. If it's not AI-complete it's probably just plain an ill-formed question. That's before we ask questions about the biases that get embedded into the misinformation detector.
This can't be fixed at Facebook scales by building a platform that so highly incentivizes low-quality content of all kinds, then trying to stop the content so late in the cycle. The entire incentive structure has to be rethought. Unfortunately, that's probably a problem that Facebook is literally incapable of solving without going out of business, because Facebook the corporate entity is that incentive structure. To fix Facebook requires Facebook to become not-Facebook.
This can't be fixed at Facebook scales by building a platform that so highly incentivizes low-quality content of all kinds, then trying to stop the content so late in the cycle. The entire incentive structure has to be rethought. Unfortunately, that's probably a problem that Facebook is literally incapable of solving without going out of business, because Facebook the corporate entity is that incentive structure. To fix Facebook requires Facebook to become not-Facebook.