The actual story is this research paper:http://arxiv.org/abs/1502.03519v1
The statistics get pretty thick. I'm not sure that I'll have the time/ability to wade through it. But from a quick skim, as far as I can tell, it still boils down to "voting" the web.
The process first extracts grammatical statements from a web page. Then the statements are statistically compared with a database of equivalent statements extracted from the entire web to decide whether or not they are "true."
It is proposed that in the future, such techniques could be used to rank pages with more "true" statements and fewer "false" statements higher.
As far as I can see, for the most part, their experimental ranking more or less lines up with conventional page-rankings. The outliers (with low "truth" score but high page rank) were a group of celebrity gossip sites and web forums.
They also identified some outliers with high "truth" scores but low page ranks, but did not try to characterize what sort of sites these were. Their point is that those sites should be ranked higher.
Anyhow, it's still only a research paper.