If you're doing this with a huge number of manually maintained rules which depend on the specifics of the ad code, you're doomed to endlessly updating signatures. That's labor-intensive. It's more effective to blacklist the ad domains and then try to infer the boundaries of the ad by looking at the DOM. That's how I did it in Ad Limiter, which filters Google search results. Google changes their ad code all the time, but Ad Limiter seldom needs to be changed.