Grokipedia is a free online encyclopedia from xAI, launched on October 27, 2025, where articles are written and checked by the Grok AI rather than by human volunteers. It has grown fast, but independent reviewers have flagged real problems with its sourcing and neutrality, so treat it as an AI draft to verify, not a settled reference.
Grokipedia is a free online encyclopedia from xAI, the company behind Grok. It launched on October 27, 2025, and its defining idea is simple: instead of human volunteers writing and arguing over entries the way they do on Wikipedia, the Grok AI writes the articles and checks them itself. Elon Musk has pitched it as a more truthful reference. Independent reviewers have pushed back hard on that claim. Both things can be read on the record, and this guide lays out what is actually known.
If you only remember one line, make it this: Grokipedia is best treated as a fast AI first draft of an encyclopedia, useful for orientation, not safe as a final citation. Below is where it came from, how it works, how it compares to Wikipedia, and how to use it without getting burned.
What Grokipedia actually is
Grokipedia is a website, grokipedia.com, that looks and reads like a general reference encyclopedia. You search a topic, you land on a long article with sections and citations, and you can follow links to related entries. On the surface it is familiar territory for anyone who has used Wikipedia.
The difference is who wrote it. Wikipedia articles are written, edited, and fought over by people, in public, with visible edit histories and talk pages. Grokipedia articles are generated by Grok, xAI's large language model, and are presented as having been fact-checked by that same model. Some entries were built from scratch by the AI. Others were adapted, in places nearly word for word, from existing Wikipedia content. Musk confirmed on October 31, 2025 that duplicating Wikipedia material was intentional at the start, framing it as a base to improve on.
That single design choice, machine authorship in place of human deliberation, is the whole story. It is why Grokipedia can grow at a speed no volunteer project could match, and it is also the root of every serious criticism aimed at it.
How Grokipedia works
The workflow is closer to a search-and-generate system than to a wiki. When you look at a Grokipedia page, you are reading output that Grok assembled from its training data and from sources it pulled in, then wrote up in encyclopedia style with inline citations.
A few mechanics matter for how much you should trust a given page:
- Articles are produced and revised by the model, not signed off by a named human expert. There is no volunteer editor whose reputation is attached to the entry.
- Readers cannot edit articles directly. If you are logged in, you can suggest an edit or report an error through a pop-up form. Those submissions are reviewed and, if accepted, applied by the Grok AI.
- The citations on a page were selected by the model. That means the quality of an article rises or falls with how good the AI was at judging its sources, which is exactly the weak point reviewers have zeroed in on.
By December 2025, according to reporting summarized on Wikipedia's own entry about Grokipedia, AI-generated edit suggestions had overtaken human submissions and made up more than three-quarters of proposed changes. In other words, the system increasingly writes and revises itself.
The version history so far
Grokipedia has shipped in visible stages, and the version numbers tell you how young the product still is.
- Version 0.1, October 27, 2025. The public launch. xAI put out an encyclopedia with, by most counts, over 800,000 articles on day one, against the roughly 7 million on English Wikipedia. The launch was rocky enough that the site struggled under early demand.
- Version 0.2, November 21, 2025. This update added multimedia features, including video summaries, and expanded the article base. It is also the period when AI self-editing began to dominate the change stream.
- Knowledge Panels, January 2026. xAI rolled out structured sidebars that put key facts about a subject in a consistent panel next to the article body, a formatting upgrade familiar to anyone who has used a search engine's info box.
The "0.x" numbering is not just branding. It is xAI signaling that this is early software. Treat feature lists and article counts as moving targets, and check the live site if a specific detail matters to you.
How big Grokipedia has become
Growth has been the headline number, and it is genuinely striking. From over 800,000 articles at launch, Grokipedia was reported to have passed 5.6 million articles by early 2026, with later third-party counts putting it above 6 million. That places it within range of the size of English Wikipedia's article base, reached in a matter of months rather than the two decades Wikipedia took.
Reader interest spiked and then cooled, which is normal for a launch this loud. Coverage noted a surge of daily visits in the first days after launch, above 460,000 on October 28, 2025, settling to a much smaller steady figure within a couple of weeks. Visits arriving from Google Search followed a similar shape: a tiny number in November 2025, a large jump into early 2026, then a fall-off by the middle of February 2026.
The lesson in those numbers is that scale and staying power are different things. Grokipedia proved it could produce entries at machine speed. Whether people keep coming back depends on whether they trust what they find, which brings us to the hard part.
Grokipedia versus Wikipedia: the real differences
It is tempting to frame this as a rivalry, and Musk has done exactly that by positioning Grokipedia against what he calls a biased Wikipedia. But the more useful comparison is structural, because the two projects are built on opposite assumptions.
- Authorship. Wikipedia is written by people. Grokipedia is written by an AI model. Human authorship is slow and messy; machine authorship is fast and uniform.
- Transparency of process. On Wikipedia, disputes happen in the open. You can read the edit history, see who changed what, and follow the argument on the talk page. On Grokipedia, the reasoning behind an article and its revisions is largely opaque, because it lives inside the model.
- Editing rights. Anyone can edit most Wikipedia pages. On Grokipedia, you can only suggest changes, and the AI decides.
- Sourcing philosophy. Wikipedia has a long, contested, but public set of rules about which sources count as reliable. Grokipedia's source selection is made by the model, and reviewers say the results are inconsistent.
A 2026 academic analysis, published in the Proceedings of the National Academy of Sciences, put the contrast bluntly: Wikipedia's openness makes its bias visible and contestable through edits and debate, while Grokipedia replaces that process with automated authorship that is harder to inspect. You do not have to accept every value judgment in that framing to see the practical point. When you cannot see how a claim got onto the page, you cannot easily check the people who put it there, because there are no people to check.
The sourcing and accuracy questions
This is where the reporting has been most pointed, and it is the main reason to be careful.
PolitiFact reviewed Grokipedia in November 2025 and found unsourced content, misleading or opinionated claims, and incorrect citations, including pages that leaned on secondhand, unattributed material such as social media clips. NBC News reported that Grokipedia cited a neo-Nazi forum dozens of times and drew on other low-credibility sources. Academic work in 2026 reported that Grokipedia articles were substantially more likely than Wikipedia's to cite sources generally regarded as unreliable, along with large counts of citations to low-credibility material and instances of the site citing itself.
Specific entries drew criticism too. Reviewers noted that the Grokipedia article on Adolf Hitler buried the Holocaust deep in the text, thousands of words in, where Wikipedia addresses it up front, and that some articles gave space to debunked claims. The Wikipedia community responded by formally deprecating Grokipedia as a source in its own articles, citing verifiability problems, circular sourcing where entries lean on other AI-generated or low-quality material, and copyright concerns tied to the copied content.
None of this means every Grokipedia page is wrong. Plenty of entries on uncontroversial topics are fine. It means the error rate is uneven and unpredictable, and it is highest exactly where accuracy matters most: contested history, health, and politics. That is the profile of a tool you check, not one you cite.
How the "fact-checking" claim holds up
Grokipedia's pitch rests on the idea that Grok checks its own work. It is worth being precise about what that can and cannot do.
An AI model checking an AI-written article is still one system judging itself against the same training data and the same source pool it drew from. If the model was confident about a shaky claim when it wrote the article, it is often confident about the same claim when it reviews it. There is no independent human with subject expertise, and no adversarial editor whose job is to challenge the framing. That is the gap reviewers keep pointing at when they find confident, well-formatted articles resting on weak citations.
The takeaway is not that AI fact-checking is worthless. It is that "fact-checked by Grok" is a much weaker guarantee than it sounds, and it is not a substitute for the primary sources themselves.
Can you contribute to Grokipedia?
If you are used to fixing a typo on Wikipedia in ten seconds, Grokipedia will feel closed. You cannot open an article and rewrite it. What you can do, once you are logged in, is submit a suggested edit or flag an error through a form. From there, the Grok AI reviews the suggestion and decides whether to fold it in.
That model has trade-offs. It blocks the vandalism and edit wars that plague open wikis, and it lets the site move quickly. But it also concentrates editorial judgment inside a single system controlled by one company, with limited outside visibility into why a correction was accepted or ignored. If you report an error and it is not fixed, there is no public talk page where you can make your case.
Licensing and the copied-content issue
Because many early Grokipedia entries were adapted from Wikipedia, licensing became an immediate flashpoint. Wikipedia content is released under Creative Commons Attribution-ShareAlike, which requires attribution and share-alike terms on derivatives. Grokipedia's own articles are offered under an X Community License that permits non-commercial and research use.
The friction is straightforward: if entries are lifted from a share-alike source, reusing and relicensing them raises attribution and compliance questions, and that was one of the grounds Wikipedia editors cited when they moved to deprecate Grokipedia as a reference. For a casual reader this is background noise. For anyone republishing or building on Grokipedia content, it is a real reason to check where an article came from before reusing it.
What Grokipedia means if you use Grok
If you already use Grok or SuperGrok, Grokipedia is part of the same world, and it is a useful window into how xAI thinks about knowledge: generate fast, present cleanly, iterate in public, and lean on the model as the arbiter. That is the same philosophy that shapes the chatbot's answers.
It also carries the same caution. A Grok answer and a Grokipedia article share a root: a confident, fluent model that is sometimes wrong in ways that are hard to spot precisely because the writing is so smooth. Reading Grokipedia critically is good practice for reading any AI output critically.
How to use Grokipedia without getting burned
A short, practical checklist:
- Use it for orientation. It is a fine way to get the shape of an unfamiliar topic quickly.
- Do not cite it as a final source. For anything you will publish, submit, or make a decision on, go to the primary source it points to and read that instead.
- Be extra skeptical on contested topics. History, health, science, and politics are where reviewers found the most trouble.
- Check the citations, not just the prose. A confident paragraph backed by a social media clip or an unreliable outlet is not evidence.
- Cross-check against Wikipedia or a primary document. If the two disagree on a factual point, that is your signal to dig.
Grokipedia is an interesting experiment in what an encyclopedia looks like when a model writes it. It is fast, it is large, and it is early. Read it the way you would read a capable but unsupervised intern's draft: worth a look, never the last word.
Questions readers ask
Is Grokipedia free to use?
Yes. Grokipedia is a free, public website at grokipedia.com. You do not need a Grok or SuperGrok subscription to read it, and there is no paywall on articles as of this writing.
Can I edit a Grokipedia article like I can on Wikipedia?
No, not directly. Readers cannot open an article and rewrite it. Logged-in users can suggest edits or report an error through a form, and those suggestions are reviewed and applied by the Grok AI rather than by a human editor.
Is Grokipedia accurate?
It is mixed. Some entries are clean and useful, but news outlets and academic reviewers have documented unsourced claims, incorrect citations, and citations to low-credibility sources. Verify anything important against a primary source before you rely on it.
Who owns and runs Grokipedia?
Grokipedia is a product of xAI, the AI company Elon Musk founded in 2023. It is built on the same Grok model family that powers the Grok chatbot and SuperGrok.