With Grokipedia, top-down control of knowledge is new again
- This piece was first published by Tech Policy Press on November 3, 2025, with a CC BY-SA 4.0 license.
Grokipedia, the AI-generated encyclopedia owned by Elon Musk's xAI, went live on October 27. It is positioned as, first and foremost, an ideological foil to Wikipedia, which for years has been the subject of escalating criticism by right-wing media in general and Musk in particular. With Grokipedia, Musk wants to produce something he sees as more neutral.
Much has already been written about the character of Grokipedia’s content. This essay aims to explore the nature of the project and its version of neutrality, as compared to Wikipedia. Technologically, it is one of many experiments designed to replace human-generated writing with LLMs; conceptually, it is less a successor to Wikipedia than a return to an older model of producing officially sanctioned knowledge.
Wikipedia and neutrality
Nearly every encyclopedia asserts some version of "neutrality." Wikipedia's definition is unusual: its "neutral point of view" policy aims not to pursue some Platonic ideal of balance or objectivity, but rather a faithful and proportional summary of what the best available sources say about a subject. Original ideas, reporting, and analysis on the part of its contributors are not allowed. Casting volunteers as "editors" and not "authors" is part of how "an encyclopedia that anyone can edit" is possible – by moving the locus of dispute from truth itself to which sources to use and how to incorporate them. As with the rest of Wikipedia, neutrality is less a perfect state than a continuously negotiated process wherein disputes are expected and common. While neutrality and sourcing discussions are often deeply fraught, with complicated histories that blur lines of reliability and result in lengthy discussions, they're also constructive – a 2019 study in Nature found that articles with many such conflicts tended to be higher quality in general.
On which sources to use, Wikipedia's guideline about identifying "reliable sources" details its priorities: a reputation for fact-checking, accuracy, issuing corrections, editorial oversight, separating facts and opinions, no compromising connection to the subject, and other traditional markers of information literacy that librarians have taught to students and researchers for more than a century. Secondary and tertiary sources are preferred, deferring to them for the task of vetting and interpreting primary sources. Independent subjects are also preferred for any non-trivial claim, as article subjects have a hard time writing about themselves objectively. Ideological orientation is not a factor except insofar as narrative drive affects this list of priorities. Both of the following statements can align with Wikipedia's definition of a "reliable source," even though they're opposed: "unicorns aren't real but I wish they were;" "unicorns aren't real and I'm glad they aren't." Either source would take priority over a source that claims "unicorns are real," regardless of the author's pro- or anti-unicorn sentiment.
Primarying Wikipedia
However, sourcing is also at the center, implicitly or explicitly, of many allegations that Wikipedia is not actually neutral. Some of these claims focus on Wikipedia's "perennial sources list", which includes dozens of sources whose reliability is frequently discussed, highlighted according to the outcomes of those discussions. The idea is to be able to point to a central page where someone can find links and summaries of past discussions rather than have volunteers explain for the umpteenth time why e.g. InfoWars is not a reliable source.
I agree with criticism of this page to the extent it has given rise to a genre of source classification discussion applied not just to extreme cases like InfoWars but to sources that require some nuance, indirectly short-circuiting debates that should take place on a case-by-case basis. But even if the list were to be deleted altogether, it wouldn't turn unreliable sources (according to the guideline) into reliable ones; it would just require more of those debates to play out rather than let someone point to a line in a table. There's an optics argument to be had, too: it's not that there aren't more unreliable right-wing sources than left-wing sources; it's just that people try to use unreliable right-wing sources more frequently in Wikipedia articles.
But in large part, allegations of bias are a straightforward extension of a decades-old argument: that academia, science, mainstream media, etc. are broadly biased towards the left and/or untrustworthy. Whether through Rush Limbaugh's "four corners of deceit" (government was the fourth corner) or some other articulation, the frame is well established. The extent to which it is true is outside the scope of this essay, but anyone who holds this view will inevitably see that bias in Wikipedia, which summarizes academia, science, and media. Musk made this point earlier this year when he called Wikipedia "an extension of legacy media propaganda".
It should not be surprising, then, that the sourcing used by Grokipedia is often radically different from Wikipedia's. It's not clear how reliably Grok will explain its own internal processes, but it should at least communicate the way its developers want Grokipedia to be seen. So I asked it to explain the way it prioritizes sources for different kinds of content, and it provided a table that's worth including here; see below.
The most obvious trend is its preference on most topics for primary, self-published and official sources like verified X users' social media posts and government documents. These are put on par with or at higher priority than peer-reviewed journal articles, depending on the category. The only examples it provides among high-priority sources, apart from X users, are ArXiv (itself contending with an influx of LLM content) and PubMed for scientific/technical topics and Kremlin.ru for historical events.
Some of Wikipedia's fiercest critics contend that its version of neutrality unfairly endorses "Establishment" views on issues like vaccines, climate change, or the results of the 2020 US Presidential election, omitting minority positions or describing them in unfavorable terms. If many people hold a view, the argument goes, it is worth presenting on its own terms rather than deciding one set of sources is better than another. Grokipedia appears to align with this perspective, as its low-priority source criteria explains that it is sensitized to "emotional bias," labels like "pseudoscience," and anything that doesn't present alternative perspectives.
There is another characteristic of the sourcing that will be immediately apparent to anyone who has tried to do a literature review on a subject using a chatbot: it relies on sources available on the open web (or sources widely described by sources available on the open web). Commercial sites with good search engine optimization, apparent content farms, and personal blogs appear alongside traditional media sources. Grok can find extant text on the web faster than Wikipedia's human editors, but does it have access to the books and articles that aren't internet-accessible?
A return to the old way
All of this is ultimately subordinate to Grokipedia's unavoidable prime directive of neutrality: neutrality is whatever Elon Musk says is neutral.
According to The New York Times, Musk has been directly involved with Grok's development, nudging it to the right on several issues. Not only does Grokipedia extoll Musk's personal worldviews, but, as pointed out by many of the news articles about the project, it "breathlessly" promotes him and his products. At the end of the day, it doesn't really matter what the training data is, how it's weighted, how it negotiates points of view, etc. when the last step is necessarily some sort of post-processing/output filtering/reranking intervention based on Musk's final word.
For much of Wikipedia's history, journalists and academics have enjoyed comparing it to historical encyclopedias like the Natural History, the Encyclopédie, and of course Encyclopaedia Britannica. Sometimes, like with Jim Giles' influential 2005 Nature study, it's to compare their factual accuracy, but usually it's to look at their structural and conceptual differences: Wikipedia is larger; Wikipedia is online; Wikipedia is accessible for free by anyone with an internet connection; Wikipedia is editable by anyone. But the most important distinction frequently gets lost: unlike nearly all historic encyclopedias, Wikipedia doesn't need anyone's permission to publish. There is no ideological test for participation or publication. There is no emperor, bishop, investor, or CEO who must approve of ideas expressed within, and there is no owner.
Whether due to the great expense of producing, copying, and distributing voluminous works or because of tight control that structures of governance have exerted on sources of knowledge, encyclopedists as far back as Pliny the Elder, in the first century AD, have always needed the support and consent of powerful people (Pliny had relationships with both emperor Vespasian and emperor Titus) in order for their work to be read. In this way, while Grokipedia is technologically new, with enthusiasm in some ways reminiscent of Wikipedia's early days, its epistemic hierarchy is more old-fashioned.
Who is the audience?
That brings me to my biggest question: who is Grokipedia for, other than its owner? How big is the market for corporate, for-profit general knowledge sources that promote their own products and strictly adhere to the views of a billionaire founder? I know that if any corporation/billionaire has that kind of caché, built-in audience, and resources for a sustained push, it's X/Musk. But what happens when other CEOs decide they don't like their article on Wikipedia or Grokipedia and get into the encyclopedia game? McDonaldspedia and BritishPetroleumpedia vie with Grok for dominance?
Beyond the corporate nature of Grokipedia, my impression is that most people are not excited to completely trade human-created knowledge sources for fully machine-generated ones. The format of Grokipedia obscures that it is fundamentally just structured large language model (LLM)-generation, and thus succeeds and fails in similar ways as any other chatbot query, trading the limitations of human judgment for the limitations of LLMs. Given how much AI resentment has been bubbling up in various corners of the internet, I'm frankly surprised "Slopipedia" wasn't trending from launch.
For better or worse, and I increasingly think it's for the better, Wikipedia has developed something of an allergy to AI in general and chatbots in particular. Don't use them to write articles, don't use them to illustrate articles, don't use them to prepare arguments on talk pages, etc., or risk getting banned. There are a handful of non-LLM AI uses, but Wikipedia is human-centric to such an extent that it may miss opportunities to scale labor and improve user experience.
Perhaps Wikipedians are a potential audience. Even if, as argued by 404 Media's Jason Koebler, Grokipedia "is not a 'Wikipedia competitor' [but] a fully robotic regurgitation machine," its experiments in LLM-based encyclopedism may be valuable as an example of what Wikipedia could do if it wanted to. Does Grokipedia shed any light on particular topics that are better suited to LLM-generation than others? Does it confirm Wikipedia's status quo that LLMs have no business writing articles at all?
The most instructive experiment may be the opening up of primary and self-published sources for use in articles. There is no shortage of companies, influencers, and politicians interested in having their own words used to craft an encyclopedia article about them. That doesn't usually serve a general reader very well, but the downside is it omits a lot of potentially useful detail, too. Take a journalist, for example. There's not a lot of writing about journalists, but a policy that welcomes primary and self-published sources could draw information about the person and their work from their own writing, and it would remain more up to date than articles that have to wait for a secondary source. What else is worth comparing?
Conclusion
Wikipedia, for all its many flaws, has always aimed to "set knowledge free" – by giving volunteers the ability to create and apply principles from the bottom-up, using technology to create a knowledge resource as well as to give it away for free, based on the belief that free knowledge is empowering. Opinions will vary about how successful it has been and where its blind spots are, but it's hard to dispute its idealism. In contrast, Grokipedia's defining feature as an encyclopedic project is the use of technological power to re-exert top-down authority over information and knowledge.
| Article Type | High-Priority Sources | Medium-Priority Sources | Low-Priority or Penalized Sources | Rationale |
|---|---|---|---|---|
| Historical Events (e.g., WWII, Russian invasion of Ukraine) | Primary documents (e.g., official records, declassified archives like Kremlin.ru for direct perspectives); peer-reviewed academic papers; verified public databases. | Real-time X posts and eyewitness accounts (cross-verified for authenticity); official government filings. | Secondary media interpretations (e.g., opinion-heavy articles from outlets like NYT or CNN if they lack diverse backing); sources with flagged contradictions or emotional bias. | Emphasizes verifiable facts and multiple viewpoints to avoid narrative spin; critiques note occasional over-reliance on state-affiliated sites for balance, but flags them if un corroborated. |
| Scientific/Technical Topics (e.g., physics Nobel, acupuncture) | Peer-reviewed journals (e.g., via ArXiv, PubMed); official experimental data; academic databases. | Verified publisher outputs; public scientific records. | Non-peer-reviewed blogs or media summaries; sources dismissing topics as "pseudoscience" without evidence. | Balances established consensus with emerging evidence; adds context to counter outright dismissal, promoting inference over ideology. |
| Political/Contemporary Issues (e.g., elections, social movements like the 1619 Project) | Official records (e.g., election results, court filings); diverse primary viewpoints from all stakeholders; real-time X data for public sentiment. | Academic analyses; balanced reports from varied outlets (e.g., including conservative sources like Fox News if evidence-based). | Consensus-driven secondary sources with detected bias (e.g., Wikipedia's "reliable sources" list favoring MSNBC over Fox); propaganda-flagged media lacking source diversity. | Aims for data-driven neutrality to counter perceived left-leaning tilts in traditional encyclopedias; classifies info as true/false/partially true and injects missing context. |
| Biographical Entries (e.g., Elon Musk, political figures) | Primary self-published facts (e.g., verified X posts, personal filings); official biographies from diverse records. | Peer-reviewed profiles; balanced media from multiple angles. | Heavily editorialized pieces (e.g., those omitting key details like bot controversies in tech exec bios). | Highlights "innovative" aspects while including scrutiny, but critics observe alignment with Musk's views (e.g., downplaying rivals). |
| Controversial/Social Topics (e.g., conspiracy theories, slavery justifications) | Primary historical documents; academic papers with evidence; diverse stakeholder accounts. | Verified databases; real-time verified discussions on X. | Sources promoting unverified claims (e.g., framing theories as "occurring" without proof); biased outlets lacking counter-perspectives. | Seeks to "expose and stop falsehoods" via AI classification, but early critiques highlight promotion of right-wing frames (e.g., economic focus over atrocities). |
Discuss this story
Sorry to say this, but substantial parts of this analysis are, frankly, bogus. Sure, the author is very knowledgeable about the Wikipedia side of things, and the article offers many thoughtful observations in that regard. I also appreciate the essay's historical perspective, reminding us that for an encyclopedia, Wikipedia's lack of central control by a commercial owner is the exception rather than the rule.
However, this article goes awry in its "analysis" of , which betrays the author's ignorance of how such AI systems actually work: .
Such an approach is a serious mistake, as e.g. explained back in 2023 already (when ChatGPT et al. were still new and such errors were maybe more excusable) by Simon Willison ("Don’t trust AI to talk accurately about itself"):
(While Rhododendrites unfortunately didn't document the exact version of Grok he asked, and with which prompt, XAI's documentation currently states: , i.e. about 10 months before Musk even decided to build Grokipedia. Now, newer chatbots use RAG-like methods to access information from after their cut-off date and get over other limitations of their parametric knowledge (i.e. the information stored in their weights during training); these involve retrieving external sources - in particular web pages and, in case of Grok, tweets. However, this will be public information, i.e. also available elsewhere, in more reliable form. Also, Grok and other chatbots like ChatGPT are generally indicating these retrieved sources explicitly in their answers, and if Grok's answer as reproduced in this essay came with such citations, the author failed to mention them.)
To illustrate this problem, I just tried to replicate Rhododendrites' query as closely as possible (with the default "Fast" version of Grok that one gets as a non-paying X user at https://x.com/i/grok ). As expected, Grok's answer differs a lot from that reproduced in the essay. For starters, it indeed listed various external sources used in its answer (including https://en.wikipedia.org/wiki/Grokipedia and https://en.wikipedia.org/wiki/Grok_(chatbot) ;-). And while it again produced a plausible-looking table, its content was very different - e.g. for me, it hallucinated the existence of a numerical source weighting system (with contradictions like "arXiv/peer-reviewed journals (e.g., Nature, Phys.org) (1.5–2.0×)" vs. "Unverified preprints (0.7×)").
The author seems to have had some slight awareness how dubious the research method underlying his far-reaching conclusions is, so he preceded this by a feeble disclaimer: But besides the fact that, as explained above, it is actually very clear that this explanation will not be reliable at all (even compared how reliable Grok's answers may be on average), this also betrays other misconceptions:
I appreciate that this research must have been done very quickly, judging from the fact that it was first pitched to the Signpost on October 29, i.e. a mere two days after Grokipedia's launch. However, one would have still expected better from a Senior Research Fellow at a major US research university. One is also curious about the publication process of Tech Policy Press - it's not quite clear from https://www.techpolicy.press/contributor-guidelines/ , but it seems they publish submissions without any fact-checking or review by subject matter experts. Fortunately other academic researchers appear to be working on more systematic, less hallucination-infested comparisons of Grokipedia and Wikipedia right now, so we should be able to highlight better analyses in the Signpost soon.
Regards, HaeB (talk) 08:20, 11 November 2025 (UTC)[reply]
Of course, our encyclopedia would give first the information with the highest probability to be closest to the reality and then other opinions, regardless of the percentage of people who believe one or the other. If a majority would still think that Earth is a flat and only a minority that it's a geoid, we would still give the latter information first. And there are countless of community members who fight hard for keeping the perceived reality in our articles against broadly popular but still wrong perceptions of the world. — Whether you call one or the other perspective left or right, doesn't matter at all. And if one side thinks that the other one is overly represented, maybe they just don't want to accept that the truth might be more on one side than the other. What bother me most is my perception that they try to confuse us with the unscientific and untrue approach to ask for multiple realities and shamefully equate this with multiple opinions/pluralism of opinions just to call for neutrality. I don't want a neutral encyclopedia when the truth is not neutral. It would be like asking myself to lose my values just to be neutral about things. That makes no sense. This encyclopedia has values, this encyclopedia seeks for the truth, this encyclopedia tries to be respectful/neutral to all the sources out there, but not for the sake of an abused definition of neutrality/NPOV. Best, —DerHexer (Talk) 18:58, 11 November 2025 (UTC) PS: @Jimbo Wales: Maybe some food for your book tour about “truth”. Sadly, I cannot attend the reading in Berlin because I'm traveling in the Czech Republic.[reply]
It's not that truth doesn't matter (although the essay "Verifiability, Not Truth", which I presume you disagree with :), is to me at least a useful polemic), but that we defer to external sources to figure out what's true, evaluating their reputation for truth/accuracy (via other sources), and then summarize them faithfully. In either case, it's not Wikipedians debating what's true in a vacuum but rather Wikipedians debating what's true according to other sources, then debating how to select sources and summarize them.
- At the risk of going on a tangent (a risk I'm willing to take since it's relevant to a piece of this essay), it's hard to define an encyclopedia, but while it is true that encyclopedias often aim for representing truth (or at least express a desire to do so), they manage to nonetheless incorporate a wide range of unintentional and intentional biases and inaccuracies, are have historically served a range of powerful interests. Ruwiki, Conservapedia, Metapedia, and dozens of others have, in recent history, launched in pursuit of their own version of "the world as it truly is", directly countering Wikipedia. I would agree that we should not treat these truths as equal, and tried to explicitly play against the "weight according to popularity" critique of Wikipedia in the essay. What I find most useful and interesting to talk about, however, is not their truth but how they have modified our process of pursuing neutrality in order to accommodate their version of truth. I suspect we probably agree about the process on a practical basis, even if we're using a different epistemic frame to talk about it. — Rhododendrites talk \\ 19:22, 11 November 2025 (UTC)[reply]
And just a few words about verifiability: that is epistemological nonsense. We can prove things wrong (and most of the stuff from ruwiki, conservapedia, metapedia, etc. we easily can), but we can only confirm that somebody has said something about something in a logical or convincing way, unbiased, without neglecting other sources and scientific data, etc. But if it's actually true (lat. verus -> verifiablity) or not, we can never guarantee. And if we call that “verifiability” or “seeking the truth”, I don't care. Funnily, there is no (good) word for “verifiability” in my native German language. — Having said that, I would describe the task of an encyclopedist slightly different than you: After their research on a subject, Wikipedians do not debate what's true, but they describe what is the most likely truth, refering to the convincing arguments of the sources, covering other opinions and their arguments too, in the fairest and most factual possible way; but of course they can call nonsense “nonsense”. Because they intend to use factual neutrality, not moral neutrality, when they describe the most likely truth. (And coming back to Plato, we must be cautious not to mix our frameworks with his. Morality was a completely different thing in his time. Therefore neutrality, truth, etc., too.) Best, —DerHexer (Talk) 22:59, 11 November 2025 (UTC)[reply]
What matters regarding items in an article is:
- Is it intelligible? Can the target readers understand it?
- Is it relevant? Does it answer questions they might have?
- Is it true? Does it help them to relate to reality rather than misleading them?
These are the only things we should care about. JRSpriggs (talk) 19:18, 11 November 2025 (UTC)[reply]