As the wave of relief washed through the Democratic Party in July following Biden’s decision to step aside, both Democrats and Republicans instantly understood that one of Harris’ biggest weaknesses was also among the most critical issues of the race — illegal immigration at the southern border. Republicans were quick to point out Harris had been appointed border czar by Biden and the failure was hers to own. Democrats denied Harris had ever been appointed “czar,” calling the claim a GOP talking point.

Voters who googled the question would likely have encountered a Wikipedia article listing presidential czars — as valuable a resource as any. Visitors who accessed the “List of executive branch czars” article on July 24 would have been informed that Kamala Harris had, indeed, served as border czar. But those who came to the page a day later, specifically after 4:02 pm Eastern Time, would have found no mention of Harris at all.

Within minutes of Harris’ removal, the article’s Talk page erupted into an edit war with numerous editors pointing out that Harris’ name had only been added to the article the day before it had been removed from it, suggesting it had been added for political purposes. The resolution to this quandary hinged on a seemingly simple, even binary question: was Harris border czar or not?

In cases of factual disagreement like this one, Wikipedia defaults to its core operating principles — including Wikipedia:No original research, which prohibits the inclusion of claims or assertions “for which no reliable, published source exists,” and Wikipedia:Verfiability, which ensures that “information comes from a reliable source.”

In this case, there was a dilemma. The media, in the wake of Harris’ overnight nomination, denied Harris had ever been border czar. “[T]he Trump campaign and Republicans have tagged Harris repeatedly with the ‘border czar’ title, which she never actually had,” wrote Stef W. Kight in Axios on July 24. Yet, a month after Biden had made a speech tasking Harris with overseeing aspects of the border crisis, Axios reported, “Harris, appointed by Biden as border czar, said she would be looking at the ‘root causes’ that drive migration.” (In March 2021, Kight herself reported that Biden had put Harris “in charge of addressing the migrant surge at the U.S.-Mexico border.”) The BBC made a similar claim the day of Biden’s speech (“Announcing Ms Harris’s appointment as his immigration czar, Mr Biden told reporters...,” the article read). Other reports from that time, including one from CNN, noted that Republicans had attempted to pin the term “czar” onto Harris’ role for political reasons, prompting the White House to push back years before Harris’ candidacy.

The debate on the article’s Talk page became heated. “Wikipedia’s editors once again showing utter contempt of history itself and an embrace of Orwellianism,” one editor wrote.

“This is just a case of contemporary politics being played with [Wikipedia] content, as ‘the border’ is the #1, 2, and 3 issue of the Trumpists. Get the banhammer ready,” said another editor.

Impassioned as it was, the Harris “czar” flap was just one skirmish amid the ceaseless battles over Wikipedia articles with even remotely political resonance.

Wikipedia articles present their subject matter with a casually authoritative, almost stolid tone. But beneath the surface lies endless argumentation played out in rounds of procedural maneuvering that would shame the most deft legislative hand. User bans, discretionary sanctions, requests for comment, arbitration cases, topic bans, page bans, deprecated sources — all encoded in a shorthand jargon — lie behind the “consensus” displayed in an article’s seemingly ripple-free surface. In a way, this arcana of behind-the-scenes conceptual machinery is Wikipedia’s most impressive feature. It’s what keeps it from grinding to a halt on infighting and intransigence.

The problem is — like with the Harris border czar reference, which is still omitted from the czar article (and will almost certainly stay that way) — the consensus it achieves often lines up with the prerogatives of the Democratic Party and the media establishment that supports it.

One of the reasons for this cuts to the very heart of how Wikipedia works. The encyclopedia is governed by a raft of policies like Wikipedia:Notability (subjects of articles should meet a threshold of notability), Wikipedia:Recentism (overdue emphasis must not be placed on recent events), and Wikipedia:Neutral Point of View (self-explanatory). None, however, play even close to the outsized role that Wikipedia:Verifiability plays, with its insistence that claims “must be attributable to a reliable published source.” The obvious question this standard raises is which sources are considered reliable. While some Wikipedia policies invite ambiguity, on this the site is clear. The Wikipedia:Reliable Sources/Perennial Sources page filters media sources into categories of “Generally reliable,” coded in a green-filled cell on the page’s table, yellow for those on which there is “No consensus,” and red for “Generally unreliable.”

The breakdown of sites filtered into each respective category is telling. The cadre of news outlets that collectively make up the mainstream media — ABC, CBS and NBC News, Associated Press, Vanity Fair, Vogue, The Atlantic, Axios, BBC, New York Times, Washington Post, NPR, Wired, CNN, AFP — are classified green for reliable. Strongly left-leaning outlets like Vox, Mother Jones, The Guardian, HuffPost and The Intercept are as well. But so are outright leftist or socialist outlets, including Jacobin, The Nation, and The Independent, as is civil rights advocacy NGO Southern Poverty Law Center.

Conservative outlets like Fox News (on politics and science), The Federalist, The Post Millennial, and Washington Free Beacon are red for generally unreliable. A lower ring of “deprecated sources,” whose use is outright prohibited, includes the Daily Mail, The Daily Caller, The Sun, NewsMax, and The Epoch Times. The Weekly Standard and Wall Street Journal (the latter of whose news pages are known for tilting more leftward than its right-of-center opinion page) are the only American conservative outlets with a green rating. Right-leaning tabloid New York Post is red; left-leaning tabloid New York Daily News is green.

While conservative American media is almost uniformly red, the same cannot be said of foreign outlets with dubious agendas. State-owned networks China Daily and Xinhua — whose purpose is to spread Chinese government propaganda to the English speaking world — get a yellow for “no consensus.” Al Jazeera, owned by Qatar, an authoritarian state, is blessed with a green reliability rating.

Regardless of its accuracy, the reality is that Reliable Sources/Perennial Sources is the fuel that keeps the engine of political Wikipedia running. With Reliable Sources/Perennial Sources at the core of how articles with political import get edited, Wikipedia is, in many ways, a wrapper for mainstream media reporting, churning news articles into neutrality-emblazoned fact. The veracity of reporting by mainstream media is not to be questioned; doing so violates an informal, but nonetheless efficacious policy broadly known as Gaslighting.

Given all this, you might think Reliable Sources/Perennial Sources is a foundational aspect of the site, ratified early on by some vote or community procedure. But you’d be wrong. While the policy of using reliable sources originated in 2005, the Reliable Sources/Perennial Sources list was created as recently as 2018. Its originator was neither a panel nor a commission of Wikipedia editors. The list was never formally adopted by the community. Rather, it was the creation of a single influential editor who, until his departure from the site in 2020, went by the handle MrX.

MrX created the list amid the heady days of Trump-related political controversies when Wikipedia’s Talk pages were marked by as much tumult as the political discourse in the broader culture. His first iteration of the list included only a single source green-coded as generally reliable: the New York Times. The Daily Mail was, already from the list’s inception, classed as red. At the same time, MrX — who, by the time he left the site was in the top 99.998 percentile of users by number of edits — was engaging in fraught debates on the site, sometimes devolving into what’s known as edit wars, on topics of extreme political sensitivity. He was highly influential in the editing of the article on Donald Trump, which (perhaps unsurprisingly) remains the first result on a Google search for Trump’s name. Between 2015 and 2020, MrX made nearly 600 edits to the “Donald Trump” article alone, not including edits to Trump-related articles.

The same year he created the Reliable Sources/Perennial page, MrX brought a wide-ranging ban against another editor, Atsme, who was also busy editing political articles (including Trump’s) albeit with a different political lens. Atsme had made clear her belief (she was identified by various users, including MrX, as a woman) that articles related to Trump were being subjected to intense political bias. In one line of argumentation, she challenged a decision by other editors to attribute racist intent to Trump in various instances. On the Talk page for an article called “Racial views of Donald Trump,” she questioned the characterization of Lindsey Graham’s reaction to Trump’s “shithole countries” comment. A group of editors described Graham’s statement as condemning Trump but, citing a CNN interview, Atsme argued Graham had in fact spoken out in defense of Trump.

Atsme similarly protested when editors sought to portray Trump as racist by referring to a reconstructed and uncorroborated conversation in a book by a former COO at one of Trump’s companies, which attributes to Trump defamatory and racist remarks against black people. Atsme argued this was an accusation of racism, since it hinged on being embraced by the media as an authentic account of the exchange. In the case of the Trump book, a priori acceptance of the media position on the statement purportedly made by Trump shifted the debate from one about whether the comments were accurate to whether they were racist. (Given their extreme nature, they obviously were.) She wrote on the article’s Talk page that “political opposition [to Trump] makes everything about racism when it isn’t, and I have cited [reliable sources] saying that very thing on this [Talk page]. We are supposed to exercise editorial judgment, not blindly repeat what journalists and pundits say, especially when there are known biases.”

Perhaps this was the point. After MrX self-instituted the Reliable Sources/Perennial Sources list, Wikipedia began anchoring assertions of factual knowledge to claims made by the mainstream media. But as the media began to widen the definition of racism to include the ever-expanding set of definitions issued by academics, BLM and the anti-racism movement — e.g. microaggressions, cultural appropriation, professionalism, structural racism, use of terms such as “master bedroom,” “blacklist” and “peanut gallery,” whiteness, and, of course, the denial of racism — Wikipedia naturally followed suit. In the case of the former Trump company COO’s book, the media took the author at face value, running articles, listicles and op-eds, including one by the book’s author, about the exchange. In the end, Wikipedia editors were satisfied with a reference to an interview Trump did with Playboy magazine to source the claim, which remains in the article.

Atsme similarly argued that, in Wikipedia’s Donald Trump article, the statement, “As president, Trump has frequently made false statements in public speeches and remarks,” had gone beyond its remit given that these were, once again, accusations of falsehoods. Atsme proposed the statement should read, “Media fact checkers have analyzed some of Trump’s statements during his first 100 days as president, and determined that he made frequent false, exaggerated or distorted claims in his public speeches and remarks.” On the Talk page, the intent of the editors on the other side of the debate was clear: Trump was by nature a liar — many contested he was a “pathological” liar — and the original statement merely reflected that objective fact (translated into the style and tone of the encyclopedia, known as WikiVoice). “[C]alling him [a pathological liar] is hardly controversial anymore, except among die hard Trump supporters with their heads in the sand, but let’s not go there 😉,” one editor quipped.

MrX’s position was defiant. Debating a re-named section title, he argued that the revised title was “not a fair representation of the overall coverage of Trump’s habit of lying. If anything, this section should be updated based on more recent tallies of his lies, and the widespread view that the lies are not simply exaggerations or ‘distorted statements’ (which is just a fancy way of saying ‘lies’).” He later underlined this sentiment writing, “I never wrote that he is a liar in the article, and I wouldn’t because it's not in keeping with encyclopedic tone. But let’s not mince words — he is most definitely a liar; a fact which is well documented in multiple reliable sources…”

On the question about whether claims about Trump’s racism should be presented as allegations, MrX fired off: “This article is not about accusations; it’s about Trump’s 45 year documented history of racially-provocative remarks and racially-motivated actions.” He bolstered his position by citing a litany of Wikipedia policies, including the policy not to question reliable sources. Other editors opposed to Atsme’s edits were just as unambiguous. One influential editor she debated responded, “Atsme, Trump has been despised ever since he slithered through the Queens-Midtown Tunnel in the late 1970’s.”

Just a few weeks after these exchanges took place, MrX brought an arbitration case against Atsme, essentially requesting she be banned from editing articles about American politics. Wikipedia guidelines for cases like this maintain that an editor who brings a case should cite two or three “diffs” — edits made or comments posted on a Talk page — to demonstrate evidence of a “smoking gun” proving a particular offense, not general opposition to another editor. MrX provided close to 100 diffs presented in a lengthy brief divided into sections. In the first section, he accused Atsme of violating WP:Gaslighting, detailing the offenses as “Repeatedly discrediting reliable sources; claiming bias and propaganda in reliable sources.”

What MrX neglected to mention, however, was that he was also the author of the gaslighting policy, which he’d created just weeks before creating Reliable Sources/Perennial Sources. Much of MrX’s case hinged not on questions of content or the merits of Atsme’s edits but on accusations of improper behavior. But on that ground, Atsme responded by showing cases where MrX had posted profanity-laced diffs, including one where he wrote to her, “I’m so fucking thoroughly sick of the constant drama and battleground bullshit on Wikipedia that I could puck [sic]. How about you and your whole busybody crew fuck right off!!”

MrX knew he was on shaky ground. By rights, a case like the one he’d presented would be brought for an “Arbitration Enforcement” (AE). But with AE cases limited to 20 diffs, while MrX had presented over 80, he had to appeal to a higher authority. In this case, an “uninvolved administrator” was brought in who issued a topic ban against Atsme nine minutes after she posted her response to MrX’s allegations, effectively silencing her from commenting further on the case.

Atsme’s American politics topic ban was later overturned on appeal to a higher body, Arbcom. But this would not be the only “T-ban” Atsme would face. She was similarly banned for posting on anti-fascism topics, where a cohort of editors had resisted efforts to show coordination behind Antifa, and had made edits to make it seem as if journalist Andy Ngo, described for a time in his Wiki entry as a “far-right” “social media activist,” had been merely “involved” in an altercation, rather than victim of an attack that left with him a brain injury.

The list of such political-reality-bending edits goes on. Eight days after Kamala Harris announced Tim Walz as her running mate, the Criticism section on Walz’s Wikipedia article was removed in its entirety. (The removal was made by an editor with administrator status who is in the top 100 list of editors by number of edits.) The article on Hunter Biden’s laptop, cast in its very title as a “controversy,” notes in the second paragraph that while the New York Post published a front-page story about the laptop in 2020, “other news outlets declined to publish the story due to concerns about provenance and suspicions of Russian disinformation.” Far from making the kind of statement of hard fact that editors had readily made regarding claims about Trump’s racism, the article fails to definitively state that the laptop is not, in fact, Russian disinformation. Instead, it offers a hedged disclaimer: “By May 2023, no evidence had publicly surfaced to support suspicions that the laptop was part of a Russian disinformation scheme.” The door was thus left open to the possibility that, four-plus years on, evidence of Russian disinformation could still arise.

The article on a possible lab origin of the virus that caused the COVID-19 pandemic, “COVID-19 lab leak theory,” mentions the variations of the term “conspiracy theory” nearly 40 times. By comparison, the article on chem trails, a canonical example of an actual conspiracy theory, uses variations of the term around 50 times. “Many scenarios proposed for a lab leak are characteristic of conspiracy theories,” the lab leak article states — once again casting aside qualifying statements to assert a bald fact. A book titled Covid Conspiracy Theories in Global Perspective written by two literature professors is cited as the source. Just one paragraph later the article asserts, “Scientists and media outlets widely dismissed [a lab origin] as a conspiracy theory.” While the former claim, that scientists as a whole “dismissed” a lab origin scenario, is demonstrably false, the latter claim about the media was true — and, as I previously reported, is precisely what distorted the scientific debate on the virus’ origin.

Instances like this are compelling, but they’re also backed up by data. Earlier this year, David Rozado, a machine learning researcher who specializes in analyzing bias, published a major study of the site. Rozado found “an average tendency in Wikipedia articles to use the names of prominent left-leaning U.S. politicians with more positive sentiment than their right-leaning counterparts.” He found the trend extends to US Supreme Court Justices and American journalists — in both groups, Wikipedia tends to employ positive sentiment in articles touching on figures on the left and more negative sentiment for figures on the right.

Most importantly, the same is true for US media outlets. Of the 10 outlets with the highest positive sentiment levels, seven are left-leaning, with NPR — whose CEO was formerly the CEO and executive director of Wikimedia Foundation (see my deep dive here) — achieving the highest sentiment. The reverse is true on the other end of the spectrum, where eight out of the 10 outlets with the most negative sentiment are conservative.

That the green-coded outlets on the Reliable Sources/Perennial Sources page align with outlets associated with the most positive sentiment should be no surprise. And perhaps this is the crux of the matter: reducing our understanding of knowledge to assertions made by a small subset of supposedly “reliable sources” — sources which, we learn every day, are not disembodied institutions but companies with interests and agendas staffed by people with biases and flaws — not only leaks bias into the encyclopedia but glazes those sources with an epoxy of false objectivity. That the list of reliable sources was created by a single individual who, it is evident, brought his own set of heavily reinforced biases to the site only makes matters worse.

There is no doubt that Wikipedia is a testament to the limitless power of collaboration and an odds-defying wonder of human achievement. The question is whether in our hyper-partisan world Wikipedia can fulfill its grand mission or if, like so many institutions whose inner dynamics overpower their founding missions, the encyclopedia is fated to achieve exactly the opposite of what its founders intended.

— Ashley Rindsberg