I think what's more likely is that they do the same thing that fantia does: if it's an image, they will insert some fingerprint data through steganography to trace which account leaks what image. Once their scrapers detect one, it flags the account that's connected to that identifier.