At Blekko.com we've looked into this, more of an a/b test driven model rather than a definitive survey, but one challenge is this, you send two pages to an unbiased observer (which is code for someone who didn't do the search, they are just looking at the search query and the results page and deciding which is 'better') the results will favor ads, but the meta issue is that the ads are curated, not algorithmic.

Specifically some person said "If someone searches for 'x,y, z' put my Ad up there" which is still much more accurate than an algorithm trying to guess. We do a game on our search page where if you put /monte at the end of your query we'll throw up the Google results, the Bing results, and the Blekko results. Then you get to pick the one that you, the searcher, thought was the best answer. We've found that well curated categories do really well in this comparison. Ads are simply a market motivated curation of Google's results.

The studies were not public and were part of research projects and relevance insights I had access to in prior positions.