What IsGrokipedia?

On 27 October 2025, a website appeared with 885,279 articles about nearly everything. It had no volunteer editors. No talk pages. No community governance of any kind. Every word had been written by a single language model called Grok, built by Elon Musk's xAI. Within ninety days it would hold more than six million entries. Within five months, Google would strip most of them from its search results. The arc of Grokipedia, from launch to correction to uneasy coexistence, tells us something important not just about AI-generated content but about what we actually expect from an encyclopedia in the first place.

Most writing about Grokipedia falls into one of two registers. Boosters call it the inevitable upgrade to a broken model: faster, broader, liberated from the editorial politics of Wikipedia's volunteer class. Critics dismiss it as a billionaire's vanity project with documented accuracy failures. Both camps miss the more interesting question, which is structural. Grokipedia is not simply a better or worse Wikipedia. It is a different kind of knowledge object entirely, one whose flaws are not bugs to be patched but consequences of a design choice that trades transparency for speed.

This article examines that choice. It draws on three independent academic studies, traffic data from Similarweb and Ahrefs, and reporting from the New York Times, the Columbia Journalism Review, the Guardian, and Wired.¹

Grokipedia is a retrieval-augmented generation (RAG) system. The Grok 4.1 large language model generates articles by pulling information from web sources, research papers, social media posts on X, and its training data, then synthesising that material into structured entries with citations.²

The decisive design choice is the editing model. Wikipedia is radically open: anyone can edit any article at any time. A community of roughly 125,000 active editors monitors changes, reverts vandalism, and argues on talk pages until consensus forms. The process is slow, contentious, and frequently maddening. It is also self-correcting in ways that matter.

Grokipedia is closed. Users cannot edit articles directly. Logged-in visitors may submit a suggestion through a feedback form, but Grok itself decides whether to implement it. The gatekeeper and the author are the same entity. There is no mechanism for productive disagreement. There is no talk page. There is no consensus process.

The Columbia Journalism Review reported in February 2026 on a Tow Center analysis finding that by December 2025, AI-generated edit suggestions had overtaken human submissions, accounting for more than three-quarters of all proposed changes.³ The encyclopedia is now mostly editing itself.

The honest answer: it depends on the subject and on whether you verify the sources.

For STEM topics, corporate filings, and fast-moving technology subjects where speed outweighs nuance, Grokipedia surfaces useful information quickly. For politically sensitive subjects, contested historical events, health information, or anything where the choice of which facts to include is itself an editorial act, independent reviewers have documented consistent problems.

The most productive approach is not to choose one platform. It is to read both. Wikipedia gives you twenty-five years of human argument: imperfect, slow, but transparent. Grokipedia gives you a language model's synthesis: fast, fluent, and opaque. The gaps between the two are where the interesting questions live.

Every article in the cont Learn section lets you toggle between Wikipedia's version and Grokipedia's AI rewrite. The differences are sometimes subtle. Sometimes they are not subtle at all. Either way, seeing both makes you a more careful reader of each.

An encyclopedia that edits itself is a genuinely new object in the world. It deserves the thing it cannot, by design, give itself: scrutiny from multiple independent perspectives, conducted in the open, with the disagreements left visible for the reader to judge.