You can't trust user reports. They will say a story is fake when they don't like it.
You can't trust some worker in Bangladesh to determine this either just as easily as you can determine if something is NSFW.
You still can't trust AI.
So how? It hires specialists in each field to determine what is real and what is not?
How the hell are they going to measure this?
If Snopes/AP tells me something is false, I'm perfectly willing to give them the benefit of the doubt. Perhaps this will change in the coming decade, but for now, this seems like a perfectly reasonable heuristic.
These are not wealthy or powerful entities.
If they're good, they get enough depth to explain what SMEs are talking about and to understand the difference between a huckster and an expert. And, amazingly enough, the SMEs themselves will disagree on fine points and big issues.
There is too much specialized information and knowledge for social media companies to be reliable arbiters of truth for any controversial topic.