When the shills in question are lazy, it can be pretty straightforward to detect, yeah. This sort of thing was very high on my list of "problems that will need mitigating if this has any chance of working" with my Suitocracy project.
The more research one does on astroturfing and the like, the more depressing it gets. Dedicated desktop VMs running different browsers and OS's, directed through different proxies to get different IPs, etc. Where social media is involved, detailed online personas are set up for each one, with reasonable histories for using later down the track.
I think the best we can realistically do is to raise the cost of successful astroturfing/shilling to the point where it might not be economical to do, at least on a meaningful scale. As others have said, community involvement and skepticism play a huge role in that.