I have to assume that there's a team of about a dozen engineers at Amazon whose entire job is solving this problem, and every so often their boss has to write a report up to SVPs about how their attempts to solve the problem are going. I imagine it involves a lot of really cool machine learning algorithms and also some simple but obvious hacks, like just blacklisting certain wide IP ranges of known bad reviewers. I imagine that the report makes a big deal about how it's successfully stopping the vast majority of these fake reviews, and it's removing more fake reviews than ever before. There's probably a series of pretty graphs that go up and to the right.

I imagine that writing this report makes their boss very nervous because I would also be nervous if I had to repeatedly tell Jeff Bezos that I was failing to stop the slowly increasing flow of people's trust away from Amazon's retail site.