> it will probably act as "quality signaling", because it is becoming more difficult every year to tell the difference

Note that this only applies to something like a translation where there's some notion of a "correct answer". For other cultural products it's irrelevant (as you say, if it quacks like a duck ...).

Quality signaling is really only necessary in situations where an upfront investment is required and any deception is only revealed sometime later upon use. Safety critical systems such as airbags are a model example of this - a counterfeit of deficient functionality won't be discovered until it deploys, which in most cases will never happen.

That said, while I certainly can't speak to business or diplomatic translations, when it comes to cultural works (ie entertainment) the appeal of machine translation to me has been gradually increasing over time as it gets better. I don't generally find localization desirable and in some cases it even leads to significant confusion when a change somehow munges important details or references. Confusion which I'm generally able to trivially resolve by referencing machine output.