I think the "preying on fish" part is key here not "scientifically observed." A captive mantis eating a fish placed in front of it is different from a wild mantis specifically hunting fish.

I wonder what reaction the authors would get if the paper they submitted was entirely based on watching Youtube?

That probably just means observed by a scientist. If I report seeing a praying mantis eat a fish, that's not as reliable as if I have a video or if a trained scientist makes the same report.

Well, the youtube video clearly"occurred under human-manipulated circumstances".

Perhaps merely “in a reproducable fashion”.