In January 2011, Google updated its auto completion feature to become more copyright friendly. In August 2012, Google made another movement in this direction. Ami Singhal, SVP Engineering, announced that the ranking algorithm will now take into account the number of valid copyright removal notices. Thus, in theory, sites with illegal content should appear lower in the pages.
Does it work? I used the same example than in related post: Black swan. The situation is better for auto completion. If you try with “Black swan pirate movie”, you do not find illegal content on the first pages. Nevertheless, if you try “Black Swan torrent”, the first page is only about illegal torrents; the first result points to the pirate bay!
By the way, when trying “Black Swan pirate movie” on Microsoft’s bing, the third proposed result points to streaming sites.
Google receives an astounding number of removal notices. According to Google’s transparency report, it received about 5,900,000 notices to remove URL for copyright infringement.
Source: http://www.google.com/transparencyreport/removals/copyright/ on 2012, September 3.
If they are all executed, it means that the execution is done automatically without a manual verifcation. There is no way to do it with systematic human supervision. May be Google uses a first step of screening with automatic fingerprinting recognition, and then manually examine the non detected ones. If somebody has a pointer that describes this process, I would love to read it. Or an educated guess?