What Is Grokipedia? Inside xAI's AI Encyclopedia vs. Wikipedia

Grokipedia is an AI-written online encyclopedia built by Elon Musk's AI company, xAI, with every article drafted and "fact-checked" by the Grok chatbot instead of human volunteers. Launched as version 0.1 in October 2025 with 885,279 articles, it was pitched as a direct challenger to Wikipedia — one Musk has long accused of ideological bias. More than a year on, Grokipedia has grown into the millions of articles, weathered an unexplained multi-month freeze in its edit pipeline, and drawn pointed criticism from researchers over copied text and unreliable sourcing. Here is what the site actually is, how it works, how big it really is, and what the evidence says about whether you can trust it.

What Is Grokipedia?

Grokipedia, at grokipedia.com, describes itself on its own homepage as "an open, comprehensive collection of all knowledge." Unlike Wikipedia, which is written and edited by a global community of human volunteers, every Grokipedia article is generated by Grok, the large language model built by xAI. Readers can't edit text directly; instead they submit a suggested correction, and Grok decides whether to accept or reject it after checking it "against sources," according to the site's own interface.

Musk first teased the project on X on September 30, 2025, writing "We are building Grokipedia @xAI," before a planned late-September launch slipped to October 27, 2025. xAI itself has not published a detailed blog post or technical paper explaining Grokipedia's methodology — most of what's known about how it works comes from the live site's interface text, Grokipedia's own self-authored entry on the topic, and independent research, rather than from an official xAI announcement.

How Grokipedia Works: From Prompt to Published Article

Grokipedia's workflow has two entry points, both run through Grok rather than a human editorial desk:

• New articles: A visitor can "name a topic and Grok writes the article, checked against sources," per the site's own suggestion prompt. Grok researches the subject, drafts the text, and publishes it without a human review step being described anywhere on the site.
• Edits to existing articles: Readers can't rewrite an article themselves. They highlight the passage they think is wrong, explain why, and optionally attach a source; Grok then evaluates the suggestion and marks it "approved," "rejected," or leaves it "in review." The site's footer sums up the model as content "written and fact-checked by Grok, corrected by the people who read it."

That is a meaningful structural difference from how Grok itself is used as a conversational assistant — readers interested in the underlying model's capabilities, pricing, and context window can see our breakdown of Grok's current model for more on that side of xAI's product line. Grokipedia layers a specific "propose, verify against sources, auto-approve" pipeline on top of that model, with no indication that a human editorial team reviews articles before or after publication the way Wikipedia's volunteer editors and administrators do.

Who Built Grokipedia — and Who Owns It Now

Grokipedia is a project of xAI, the AI company Musk founded in 2023 and best known for the Grok family of chatbots. In February 2026, Musk announced that SpaceX had acquired xAI in a share-swap deal, making xAI — and by extension Grokipedia — a subsidiary of SpaceX rather than an independent company. That corporate change hasn't altered how Grokipedia functions day to day: it's still a Grok-driven, xAI-branded product, just one now sitting inside Musk's larger SpaceX corporate structure.

Grokipedia's Scale and Version History

Grokipedia's article count has been reported wildly differently depending on when you check and which source you trust — and the site's own numbers don't always agree with each other. Here is the version history as best it can be reconstructed, with each figure attributed to its moment in time:

That last discrepancy is worth sitting with. Visiting grokipedia.com directly on October 25, 2026, the homepage footer displayed "0 articles" while a separate ticker at the top of the page showed roughly 1.23 million "approved edits" — and Grokipedia's own article about itself, last touched the day before, still listed 6,092,140 articles unchanged since January 2026. Nothing on the site explains the "0 articles" figure; it may be a display bug introduced with the September 2026 v0.3 redesign rather than evidence the catalog shrank. Either way, it's a useful reminder that even Grokipedia's self-reported numbers should be treated as a snapshot, not a verified fact, until confirmed independently.

For scale, English Wikipedia's own statistics page lists more than 7.2 million articles — meaning that even at Grokipedia's highest reported count, Wikipedia's human-written catalog remains somewhat larger, built up over nearly 25 years rather than one.

Grokipedia vs. Wikipedia: How the Two Differ

The two sites look similar — long-form articles, infoboxes, citations — but the production model is fundamentally different:

• Who writes it: Wikipedia articles are written and revised by volunteer editors who debate changes on talk pages and can be blocked for misconduct. Every Grokipedia article is drafted by Grok; there is no visible community discussion layer, and readers can only suggest changes, not make them directly.
• Review: Wikipedia relies on peer review, citation requirements enforced by other editors, and a public edit history. Grokipedia relies on Grok itself judging whether a suggested correction is accurate, with the decision logged but not accompanied by a visible diff on most entries.
• Transparency: A Wikimedia Foundation spokesperson, responding to Grokipedia's launch, said the group was assessing the project but that Wikipedia's strength lies in its transparency and "human-created knowledge" — a direct contrast Wikipedia has drawn with Grokipedia's AI-authored model.
• Underlying content: Academic researchers found that much of Grokipedia's content is "highly derivative of Wikipedia," including large sections reproduced under Wikipedia's Creative Commons license, even as other sections diverge sharply, particularly on political topics.

If you're weighing AI tools generally against each other rather than against Wikipedia specifically, our comparison of Claude, Grok, and ChatGPT as assistants covers the chatbot side of that question, while Grokipedia is really a separate product built on top of one of those models.

Is Grokipedia Accurate? What Researchers Found

The most rigorous public assessment so far comes from Cornell Tech researchers Harold Triedman and Alexios Mantzarlis, whose paper "What did Elon change? A comprehensive analysis of Grokipedia" was posted to arXiv on November 12, 2025. Comparing Grokipedia's corpus against a full Wikipedia dump, they found:

• Grokipedia articles are substantially longer than their Wikipedia counterparts on the same topics.
• Much of Grokipedia's content — including text nominally under Creative Commons licensing — is "highly derivative of Wikipedia," i.e. closely copied rather than independently written.
• Where Grokipedia does diverge from Wikipedia, it cites far more sources that English Wikipedia's own editorial community classifies as "generally unreliable" or blacklisted — the researchers measured non-derivative Grokipedia articles as 3.2 times more likely to cite such sources than Wikipedia articles, including citations to sites like Stormfront and Infowars.
• Divergence and error rates were highest specifically on articles about elected officials and other politically contested topics — exactly the subject areas Musk has said Wikipedia gets wrong.

Specific errors have also surfaced in individual articles flagged by journalists and by Wikipedia co-founders themselves. Wikipedia co-founder Larry Sanger reviewed Grokipedia's entry on himself shortly after launch and, writing on his own site in October 2025, described the prose as readable but "often insipid pablum," and traced some inaccuracies back to Grok relying on flawed source material rather than reasoning errors of its own. Sanger said the project might eventually improve past Wikipedia precisely because an AI reviewer could reject obviously bad edits that a human editor might miss — but stressed that at "v0.1," it wasn't there yet. Wikipedia co-founder Jimmy Wales, separately, said he expected "a lot of errors" from any encyclopedia written by current-generation language models, arguing LLMs "are going to make massive errors" on anything requiring nuance.

The Bias Debate: Claims From Both Sides

Bias accusations run in both directions. Musk has repeatedly described Wikipedia as, in his words, controlled by "far-left activists" with "extremely left-biased" editorial control, and framed Grokipedia as a correction to that — part of why he delayed the October 2025 launch, saying he wanted time to "purge out the propaganda" before release. Grokipedia's marketing has leaned into being an "anti-woke" alternative, and some early reviewers found its tone on politically charged topics noticeably different from Wikipedia's.

Critics make the opposite case: that Grokipedia's sourcing choices tilt it toward fringe and partisan material. The Cornell Tech researchers specifically found elevated citation rates to far-right, state-media, and conspiracy-oriented sources on Grokipedia's non-Wikipedia-derived content, and other reporting has pointed to individual entries — such as one on a long-debunked "Clinton body count" conspiracy theory — citing outlets like Infowars as if they were reliable. Neither side's claim is fully resolved by the available evidence: Grokipedia clearly diverges from Wikipedia's editorial norms on contested political subjects, but whether that divergence reflects correcting left-leaning bias or introducing a different bias is, unsurprisingly, contested along the same political lines as the underlying dispute.

The Long Pause: Why Grokipedia Stopped Updating

One of the more concrete, verifiable findings about Grokipedia isn't about bias at all — it's about neglect. Renée DiResta of Georgetown's McCourt School and researcher Ronald Robertson published an analysis, discussed on the Lawfare Daily podcast in August 2026, finding that Grokipedia's edit-review pipeline appears to have frozen around April 24, 2026. Their research, built in part by reconstructing site activity from Grokipedia's own search typeahead since the site publishes no activity rankings itself, found more than 13,000 user-submitted suggestions stuck in "in review" status with zero resolved in the prior three months, and no public announcement from xAI explaining the halt. They also identified a mass rewrite on March 14, 2026 that broke the technical links between suggested edits and the text they referenced, retroactively mislabeling some previously accepted edits as "rejected."

Grokipedia's own self-authored article on the topic corroborates the broad timeline, describing updates as "paused from about April 24, 2026, until late September 2026." That gap lines up with the v0.3 relaunch on September 30, 2026 — suggesting the freeze may have been tied to a larger rebuild rather than abandonment, though xAI has not published an explanation either way.

How to Access and Search Grokipedia

Grokipedia is free to read at grokipedia.com with no account required. A search page lets you look up a topic directly, a "Suggest an article" flow lets signed-in users request new entries, and a "Watch live" feed shows edits as Grok processes them in close to real time (when the pipeline is actually running). Submitting a correction or requesting a new article requires signing in; unsigned drafts are saved locally to your device rather than submitted. There is no official mobile app at this writing — the site is web-only. Readers who want an AI-assisted alternative for open-ended research rather than a standalone encyclopedia might also look at how AI search tools compare in our rundown of Perplexity, ChatGPT, and Claude as AI search assistants, which serve a related but distinct purpose from a browsable encyclopedia like Grokipedia or Wikipedia.

Bottom Line: Should You Trust Grokipedia?

Grokipedia is a real, large, and fast-growing reference product — by its own count, in the low millions of articles and over a million approved edits as of late October 2026 — built entirely by an AI model with no visible human editorial layer. That makes it fast to generate new content and, according to its backers, potentially less susceptible to the kind of slow-moving editorial consensus Musk criticizes in Wikipedia. It does not yet make it reliable in the way Wikipedia has become reliable: independent researchers have documented substantial copying from the very source it aims to replace, measurably higher rates of citations to low-quality sources on the topics where it diverges from Wikipedia, and a multi-month stretch where the entire correction pipeline simply stopped working without explanation. For now, Grokipedia is worth treating the way you'd treat any single AI chatbot's unverified answer — a useful starting point, not a citable authority — until a site update, an independent audit, or xAI's own disclosures narrow the considerable gap between what it claims about itself and what outside researchers have been able to verify.

Written byTheo Park

Theo Park runs the AI desk at Pandromeda. He follows model launches from the frontier labs and the open-weight community, tracks the assistants and developer tools built on them, and explains what each release changes on pricing, capability and safety. His reporting leans on primary sources: model cards, technical reports, API documentation and the companies' own announcements.

More from Theo ParkAll AI coverage