Reverse Search

I don't know if this has already been done. But wouldn't it be interesting if you could specify a few links and see what string search you would have had to do to get those links with reasonable ranks?

I think it would be cool. Can see how sites are similar, if they are at all. It could even be used to cluster links together with phrased based matching, I wonder if the solution to the reverse search problem could be done in a different way that data mining algs do phrase-based document clustering...