You can tell this site has no credibility based on the other articles on the site. I think the trick with a machine learning approach to identifying these site

You can tell this site has no credibility based on the other articles on the site. I think the trick with a machine learning approach to identifying these sites is not letting the model fixate on particular words (e.g. Obama or Moon Landing could be terms with the right dataset; though if they occur together maybe that's a good indicator after all...)