It's pretty easy to spot fake reviews. The tacticts are pretty obvious. I'd bet basic NLP and pattern-matching would work to catch a huge percentage of fake-reviews if it were something Amazon really cared about doing.

Is the reviewer saying things of substance, how many other reviews have they left and over how many years, how many things have they actually bought on the store and at what dollar amounts, are their other reviews a mix of positive/negative, are the types of products distinct, etc. Score each of these and remove/shadow-ban reviews that don't meet an authenticity metric.

If there are real consequences for getting caught buying fake reviews (e.g. banning from selling or fines), it could really restore some order to the chaos. (Then you get the inverse problem of competitors buying fake reviews trying to get the competition banned or fined. This is second-order to actually removing the fake reviews in the first place though.)