You shouldn't trust Wikipedia

It's a polluted, biased information ecosystem also training your AI

I previously wrote what turned out to be a very popular piece about Reddit’s descent into madness, how a platform that once hosted genuinely diverse, chaotic, interesting discourse was slowly captured by a small class of ideologically-motivated moderators until it became a monoculture. Since that story resonated with many readers, today I want to talk about how something similar is happening on Wikipedia, far more subversively, with even more dangerous consequences.

Many people don’t know any better and treat Wikipedia as a “neutral” starting point. Reporters cite it, students build papers from it, it’s become ubiquitous in our culture. And increasingly, this is the part that should most concern you, Wikipedia is one of the most influential information sources represented in many AI training datasets (concerningly, Reddit is too). It’s now far more than simply an encyclopedia “anyone can edit” (casual contributors will find their edits reverted quickly, even if true), it’s become operating infrastructure for how internet users (try to) understand reality. Increasingly they do not.

Which is why two recent investigations deserve your attention, and there could easily be 100s if not 1,000s more if people bothered to look.

Ashley Rindsberg at Pirate Wires published a detailed investigation into how a coordinated group of roughly 40 Wikipedia editors systematically worked to reshape the Israel-Palestine conflict narrative: delegitimizing Israel, softening the portrayal of Hamas, and pushing fringe academic positions into the mainstream of articles read by millions. Six weeks after October 7, one of these editors successfully removed mention of Hamas’ 1988 charter (the one calling for the killing of Jews and destruction of Israel) from the Hamas article entirely. A separate group, Tech For Palestine, ran a parallel campaign coordinating edits across an 8,000-member Discord server in open violation of Wikipedia’s own policies. When a blogger discovered what they were doing, they panicked, deleted their talk pages and sandboxes, and one editor wiped her entire chat history. That’s not the behavior of people confident they were doing something legitimate.

The Cipher Brief just went further, examining how anonymous Wikipedia editing is now directly shaping AI systems and global narratives, including an investigation into how Al Jazeera’s Wikipedia entry was systematically reshaped, with downstream effects on how every search engine and LLM now characterizes Qatar’s state-funded media network. By the way, this problem isn’t confined to one conflict or one topic.

In addition to foreign adversaries shaping narratives and perceptions, the domestic political bias in Wikipedia is also widespread. A study covered by National Review found evidence suggesting what critics have argued for years: Wikipedia skews demonstrably left across political topics. Libertarian writer John Stossel documented the mechanisms: the editor demographics, the sourcing hierarchies that privilege certain outlets over others, the way “reliable sources” gets defined in ways that systematically exclude center-right perspectives. The bias isn’t accidental. It’s baked into who edits, what counts as a credible citation, and who has the time and motivation to fight edit wars until the other side gives up.

Now consider who ran Wikipedia before becoming CEO of NPR. Katherine Maher led the Wikimedia Foundation, then went on to run public broadcasting’s flagship network, which if you’ve listened to has a very clear ideological slant. Anyway, in a 2022 TED Talk she gave while running Wikipedia, she said this: “Perhaps for our most tricky disagreements, seeking the truth, and seeking to convince others of the truth, might not be the right place to start. In fact, our reverence for the truth might be a distraction that’s getting in the way of finding common ground and getting things done.”

Read that again if you didn’t carefully enough, or better yet watch her say it. The person running the world’s most-cited reference site believed that truth was getting in the way. This is not some peripheral figure, nor a fringe view held by a lone editor. This is the institutional philosophy, stated plainly, by the person at the top. It explains a lot of what’s wrong with our digital information ecosystems, which are seriously polluted, as many others editing the web hold this same perspective. If you don’t yet see these people as the enemies of our civilization they are, you should start to before it’s too late. To my dedicated liberal readers, I find nothing about these people actually liberal, they are not on your side and represent something else.

There’s more subtle things that are awful happening here too I personally have noticed. There's a thriving academic industry (that would applaud Maher’s post-truth ideals) dedicated to finding historical figures "problematic," and its output flows directly into Wikipedia as citable sources, which then flows into AI. For a project I was interested in a quote from William Osler, the father of modern medicine, who invented the residency program and transformed how physicians are trained worldwide, and then went to look him up. His Wikipedia page calls him a "renowned practical joker" in one paragraph, then in another cites a 2020 grievance paper that uses his satirical fictional alter ego (a fake character invented specifically to prank colleagues, whose writings were never meant to be published) as evidence of racism. The same page, the same man. Wikipedia isn't noticing the contradiction because it's not trying to give you a coherent portrait; it's a document with many authors and many agendas, and the grievance authors are always the most motivated ones in the room and get consistently high placement. One citation at a time, great figures get quietly stamped "problematic" in the infrastructure that trains your AI. Osler quite literally willed his personal library to McGill University and spent his final years building the institutions that train physicians to this day, the idea that future doctors should encounter his name bracketed by a caveat about a joke he played on a colleague in 1884 is honestly vandalism.

And it’s not just how Wikipedia treats the historical dead. It’s also how they treat living, demonstrably notable people who don’t fit whatever invisible checklist their editor class has decided matters. A friend of mine Barry Schwartz has covered search engine optimization and Google’s algorithm changes longer and more consistently than virtually anyone alive. He founded Search Engine Roundtable, has been cited as a primary source by journalists, researchers, and industry professionals for over two decades, and is widely considered one of the most important chroniclers of the SEO industry’s entire history (this is a multi billion dollar sector). His Wikipedia page was deleted anyway, not once, not twice, but three times. If you go read the deletion discussion, you will spend the next twenty minutes shaking your head. The editors debating his “notability” write with the confidence of people who have never heard of him, which is itself the tell. There are Wikipedia pages for moderately trafficked food blogs and mid-tier local politicians, but the most prolific SEO journalist of his generation doesn’t clear the bar. The criteria aren’t applied consistently because the criteria aren’t really the point, the point is who controls the room. Danny Sullivan, who helped build the search marketing industry alongside Barry and works at Google, called it out plainly: Barry has made a recognized contribution to the historical record of search longer than almost anyone. That apparently isn’t enough. What’s particularly rich is a possible outcome here, if someone wrote a definitive book on the history of SEO, Barry would get a page the next week. The encyclopedia would credit the book and the book would have gotten half its material from Barry. You see how absurd all this is, I hope.

The uncomfortable truth about Wikipedia is the same as the uncomfortable truth about Reddit, about academia, and about many previously prestige information institutions: they did not become captured through conspiracy. They became captured through neglect. Organized groups with strong ideological commitments and moderators on egotistical power trips simply showed up, consistently, until they set the defaults. The rest of us naively assumed the system was self-correcting. It isn’t.

Use Wikipedia to find out the year a movie was released or about a breed of dog. For anything touching politics, history, foreign policy, so-called ‘notable’ figures, or contested science, treat it the way you’d treat any other source with a known point of view. Read it, note who wrote it, check what it’s citing, and don’t let it be the last stop. It’s not special, you should be skeptical, and you definitely shouldn’t blindly trust it. With certainty we can say Wikipedia is a reason why many people are misinformed about the world, and their management would consider that of greater importance than truth (their own words).

The encyclopedia that trained your AI has an editor class with a subversive agenda, and it’s one the majority of Americans don’t agree with. What we really want is Veritas, but you won’t find it on many parts of that website.