A patent was recently granted to Microsoft for “Generating content to satisfy underserved search queries” invented by Mark Looi. To quote Bill, I have the following abstract:
I think what the search engines might do is use their aggregated data acquired over the years to form relationships between the different under-served search queries to try and serve relevant online resources.
Google already suggest via Google Suggest which keywords you should be searching for so this is a logical extension for both search engines to suggest which search results you should be getting.
This may include the knowledge graph as Bill suggests. If knowledge bases are indeed used then this could reduce the amount of long tail traffic available to sites of small but growing reputation as Google and Bing may simply re-categorise these under-served queries for normal or hyper served queries thereby shutting out information on a body of knowledge that is small. Perhaps the under served queries are un-deserved.