The first level of online censorship happens without the victims even knowing it’s happening. Tweets, posts, articles, videos, comments, and websites of political content are all uploaded without resistance. But they aren’t seen, aren’t suggested, and are swiftly buried under a pile of competing content.

After the 2016 election delivered a result that shocked Silicon Valley, the tech giants one by one came out with statements saying that as a result of the “misinformation” circulating around the election they were adjusting the algorithms that rank their content.

As Google’s Senior Vice President of Search Ben Gomes said after the 2016 election, “We’ve adjusted our signals to help surface more authoritative pages and demote low-quality content.” In addition, Gomes said it updated guidelines for “raters to appropriately flag, which can include misleading information, unexpected offensive results, hoaxes and unsupported conspiracy theories.”

There is a lot of content online, and if losing some disreputable content in the shuffle were a neutral process, that would be one thing. But Google, Twitter, Apple and other tech giants are not disinterested curators of online content. About 95 percent of all political donations from technology company employees went to Hillary Clinton during the 2016 election, according to the political money-tracking firm Crowdpac. A video leaked from Google’s first all-staff meeting after the election makes this even clearer, as Co-Founder Sergey Brin told his staff the election result was “deeply offensive,” and Financial Chief Ruth Porat struggled through tears to vow that Google would “use the great strength and resources and reach we have to continue to advance really important values.”

I doubt executives from Google and other Silicon Valley companies would see eye-to-eye with the majority of the country about which values are really important, what is or isn’t offensive, what is a hoax or a conspiracy theory, and the difference between authoritative and fake news. In fact, a Pew Research Center study conducted last year confirmed that “72 [percent] of the public thinks it likely that social media platforms actively censor political views that those companies find objectionable.” That belief is held across the political spectrum, though it’s even stronger among conservatives, with 85 percent of “Republicans and Republican-leaning” independents saying they believe the tech giants are censoring political speech.

It turns out that distrust is well-founded. In the months leading up to the 2016 election, research scientist Robert Epstein led a team that objectively determined that the top ten results given on the first page of thousands of Google searches about the election on average favored Clinton over Donald Trump. He later estimated this may have shifted two million to three million undecided voter votes to Clinton over Trump. In June, he told Fox News host Tucker Carlson that manipulated search results could swing as many as 15 million votes to Democrats in the 2020 election.

Epstein has gone on to conduct peer-reviewed research suggesting that Google is manipulating opinions by determining which search terms it suggests from the very first character typed, and those that flash at you as you continue typing.

Google isn’t the only one doing it. A July 2018 report from VICE News showed that, “Twitter’s drop-down search section was not showing the profiles of some prominent Republicans in what amounted to a “shadow ban.” When users began typing in the names of Republican National Committee Chair Ronna McDaniel and others, profiles of less well-known people with fewer followers were suggested instead. Democratic National Committee Chair Tom Perez and other Democrats weren’t having the same problems. Twitter later fixed the problem and has repeated its testimony to Congress that it doesn’t make ranking judgements based on political views. Product Lead Kayvon Beykpour said later the problem was due to Twitter’s use of “behavioral signals and machine learning to reduce people’s ability to detract from healthy public conversation,” which mistakenly marked McDaniel’s and others’ accounts as problematic.

It was the machine, you see. Pay no attention to the man behind the curtain.

Intentional or not, Epstein’s research shows that these search-result manipulations make a huge impact. Favoring one politician or political view in a search result over another “can easily shift the voting preferences of undecided voters by 20 percent or more—up to 80 percent in some demographic groups,” Epstein wrote in USA Today.

There is yet no way to escape Google political bias, as its influence is enormous. It controls an 88.5 percent share of the global search market and its YouTube subsidiary controls 90 percent of the online video market, according to Statista.

Which brings us to demonetization, the second level of online censorship.

YouTube video creators have long been familiar with the term, in which videos they create with problematic content are denied advertising revenue. On YouTube and other platforms demonetization is an algorithmic function, in which a list of topics and words once uttered aloud will prevent video creators from earning a living. It functions as a machine-controlled Overton window that strictly limits the terms of acceptable debate. The result is that video channels that deal with sensitive political subjects can find their entire catalogue demonetized. This happened with several channels focused on gun ownership and Second Amendment rights, as well as channels that break down video of police shooting incidents in an effort to combat misinformation spread by activist groups like Black Lives Matter.

In June, YouTube demonetized all content on conservative comedian Steven Crowder’s highly popular channel, after he did an impression of gay Vox Media host Carlos Maza’s lisping speaking style. Maza called it homophobic harassment, and YouTube called it an “ongoing pattern of egregious behavior.” Crowder could receive ad revenue again, YouTube said, but first “all relevant issues with the channel need to be addressed.”

The final step is deplatforming, which is to say, outright banning. The list of conservatives who have been deplatformed from social media applications is long and growing longer every day. From right-wing provocateurs Milo Yiannopoulos, Alex Jones, and Gavin McInnes, to center-left Brexit proponent Carl Benjamin.

All of these and more have been labelled by Big Tech as too dangerous to be allowed to speak. Meanwhile, leftist celebrities who threaten violence toward conservatives, and Antifa members who act it out, continue to keep their accounts.

If this catalogue of online censorship is depressing despite its brevity, consider that the first step toward a solution is to understand that there is a problem. What can be done about it at this point is not totally clear.

In the next article in this censorship series, Chronicles’ Legal Affairs Editor Stephen B. Presser suggests one path forward.