The idea is to have a program that scans web pages and classifies the major blocks of screen space. I'm not sure how this could be done without manual labor at the moment that's why it's half-baked.
So it would take a webpage, and highlight all adspace blue. Then highlight all legal disclaimer boilerplate
stuff yellow. Then highlight all self-referential links and logos orange. Then highlight all completely non-relevant content another color etc. (who needs the top 10 music downloads on a weather site anyway?)
The end result would be some (albeit small) percentage of actual original relevant content.
One would presumably use this information comparatively for deciding which websites to visit frequently. I.E. if Yahoo! has more original content than CNN, then I would choose the one that is less cluttered...
It could periodically send awards to the most clutter-saturated site, and the most garbage free site etc.