It's anyone's guess what will happen in the long run (or rather, the medium run) but for now the approach that seems to be working is for users to flag the comments that they guess are bot-generated. That's actually the approach we've always taken—it's just that the bots, as everyone knows, have taken a quantum leap recently.