I saw this write up about Cornell coming up with a way to detect fake online reviews with about 90% accuracy. It seems like a simpler problem than detecting spam, though I don’t know if that is true or not. It’s worth opening up the link to see a simple test, two reviews, read them and pick the one you think is a fake. I tried and picked the wrong one.
It doesn’t go into the “how” too much, but a few notes that give you an idea. The challenge is that the fakers will soon enough be getting smarter and trickier, an arms race that is hard to win but worth fighting nonetheless.