Perhaps they are cross-matching the data as I've described above?
WordTracker have this to say in their manual about their data sources:
"We get our data from metacrawlers, rather than the search engines themselves. After all, the metacrawlers contain the results from the search engines.
To be fair we do make an assumption. We assume that people will search for the same things regardless of whether they use a search engine like Yahoo (
www.yahoo.com) or a metacrawler like Metacrawler. The argument can be raised that many new people who have just joined the internet wouldn't know a metacrawler if it jumped out at them. But then... it works both ways. There are many portals which use metacrawlers for searching the web (
www.cnet.com for example).
We believe that a user (especially a new user) would see a search box as a search box. It matters not the technology underneath. Also remember that a metacrawler uses the major search engines for its results!
In any case, we have found a great deal of similarity between the results from search engines and metacrawlers. Concrete examples can be found by comparing our top 100 lists to those of Altavista, Lycos, Searchterms.com, Searchwords.com and an analysis of 1 billion top Altavista searches from 1999.
The other great thing about metacrawler results is that we do not have to contend with the skew from people using software robots checking keyword positions. All in all, it works out pretty well!"