I found the silent videos much harder, though more in the direction that I marked a lot of them as fake when they were in fact real.
I didn't get the reasoning behind the text-only examples at all. It's impossible to verify a quote just by looking at it. We know this for several centuries - that's why academia makes such a big deal of tracking sources. So what were they trying to do here?
In general, I think the performance results here might be a bit too optimistic: It's one thing to spot a fake if you are told to specifically look for them. However, if one of those were just casually embedded in some news report while the attention of the audience is somewhere completely different, I doubt we could spot them as easily. Even moreso if the fake isn't of the two most well-known faces of the western world.