This 30+ page technical slide deck explains how Google is able to leverage contemporary data science and machine learning to classify web sites, visitors, and advertisers based on contextual analysis. It also explains how authoritative journalists, raters, blue checks, and narrative owners generate hit pieces, articles, and content designed to exploit Google's machine learning to attack arbitrary opponents.
Google has to continuously update their contextual associations to ensure their product can connect advertisers to potential consumers. Their systems are continuously trained by a large volume of daily content changes. Google provides "authoritativeness" ratings to content associated with to "authoritative" people and organizations, allowing them give them greater influence over the contextual associations Google forms. If these authoritative actors all create their content in unison to make sure Google's machine learning establishes specific contextual associations, they can engage in vanishing gradient stuffing, which is a clever way to exploit an underlying property of machine learning to hide non-authoritative content from control search engine behavior, advertising and commercial opportunities, and future content creation.
You can download this 30+ page technical slide deck to understand precisely how journalists are actively engaging in political censorship.
Much of this research was derived from Zach Vorhies's leaks. You can quickly search those leaks at https://google-leak.surge.sh.