Having identified the ambiguities of place names and human expressions of geographic references rooted in the theory of naive geography it is clear that georeferencing documents based on place names and phrases has a certain uncertainty and that the geographic footprint of a document can be seen as correct only with a certain probability.
There are many alternative approaches to georeferencing, including identification of server IP, analysis of linking topology to find out if a certain page has links to information about certain areas, analysis of server logs to find out the location of users that access the information (people tend to access local pages more often then other pages) and more, but such methods really georeference the files/documents carrying the information rather than the information itself.