You can use the criteria of inauthentic behavior. You can use the criteria of what we used to use about spams and scams on the Internet, by people voluntarily flagging it. You can use the criteria of paid sponsorships, and so on. Each criteria would yield a different picture.