> How would Google know which postings are fraudulent and which aren't?

There's a very very simple algorithm for this.

    Do I have a direct business relationship with this company?
     |
     |-> Yes -> take orders on their behalf
     |-> No  -> Do not take orders on their behalf