It's probably pretty simple... a "brilliant move" is a blunder at low depth, maybe depth 10 or something (which is really low for an engine) but at a higher depth (let's say 15 or 20) it's the best move (which is completely possible for even low rated humans now and then).
All the "brilliant" moves I've seen, including the one I played in one of my games, are unremarkable.
Several different people have asked this question on this forum, but for some reason no one gave even a half decent answer that I could see. So I did a little digging and thinking and found what I am fairly certain is the correct answer:
First, I should say that someone also actually answered this on Youtube, minute 3:45 to minute 5:00 of the following video:
https://www.youtube.com/watch?v=Zes3IaHu7Sw
That said, the answer in the above video was pretty poorly worded and hard to understand.
The more clear/correct answer appears to be that a “brilliant move” is a “best move” which the engine thinks will be very difficult / non-intuitive or impossible for your opponent to counter. i.e. either it will result in a major shift of score on the black vs. white heuristics (shift to the negative if you are, say, black) if you continue on a particular pattern, regardless of what your opponent does, or, alternatively, there is a move your opponent can do to counter the move you have done but the reason for that countermove being much better than any other move is extremely “non-obvious”, at least according to the thinking of the chess engine. How exactly the chess.com chess engine decides what a non-obvious move is and how accurate that is a question for a programmer more than a chess player—don’t ask me!