We could test this. Take a few hundred search queries, strip the pages that display Google ads, and see if the remainder of the search result is better or worse.
We'd need to get some humans in to rank the results, but that's not a big problem. "How well does this web page answer this query, on a scale of 1-10?"
With a collection of ranked pages, we can answer other questions as well. I'd be interested in running the same test but for google analytics, not google ads, as I think there might be a misaligned incentive there too.
It's worth bearing in mind that the stackoverflow clones may actually answer the query just as well as the original site - that is, it might be our definition of "a good result" that's out of whack (because we have an unnecessary bias towards the original source). I doubt this, but again it's something that's testable.
