Even the honest ones are forced to get fake reviews, because their competitor is winning with fake reviews. This is the cost of replacing humans with algorithms, before it is on par in accuracy.
The problem here is not replacing humans with algorithms, it's with replacing trusted humans with arbitrary humans (and not even checking whether one human is pretending to be many humans). This is the same problem with online forums and comment sections. We need a way to build a web of trust, so I can tap my network for reviews, but have a way to blacklist the shills who pop up.
There's no reason the "average" from the reviews presented need to count them all... the search ratings could easily exclude reviews by those that didn't have a confirmed purchase.