one year on
Musk launches Grokipedia, an AI-written encyclopedia, after hours-long outage
xAI's Wikipedia rival goes live with 885,000 articles and a suggest-an-edit tool that users still can’t actually edit through — and disclaimers admitting some pages are adapted from Wikipedia itself.
Originally published Published on this site
xAI launched Grokipedia on Monday, an AI-generated encyclopedia the company bills as an improved, less biased alternative to Wikipedia. The site went down for a few hours on launch day, according to web archive records, before returning with a counter showing 885,279 articles. English Wikipedia lists more than 7 million.
Grokipedia is labeled version 0.1. Musk, who announced the project last month after a suggestion from White House AI and crypto czar David Sacks, said on X that a v1.0 would be “10X better,” while calling the current build already better than Wikipedia. Entries carry headings, subheadings and citations, and pages state they were fact-checked by Grok.
Some entries also carry a footer reading, “The content is adapted from Wikipedia, licensed under Creative Commons Attribution-ShareAlike 4.0 License.” On pages including the MacBook Air, PlayStation 5 and Lincoln Mark VIII, that notice sits atop text nearly identical to Wikipedia’s. Entries for Trump and Musk omit subjects covered at length on Wikipedia, among them the Qatari jet, a Trump-themed meme coin, and Musk’s January rally gesture.
Readers cannot edit pages; a pop-up form lets them suggest edits/report wrong information. Wikimedia’s foundation said it was still figuring out how the site works.
The record
Said the foundation is still working out how Grokipedia functions, that alternative encyclopedias have come and gone without disturbing its mission, and that Wikipedia's knowledge is and will remain human — adding that even Grokipedia depends on Wikipedia existing.
The Wikipedia co-founder, long a critic of the project's editorial leadership, posted a lengthy thread after launch cataloguing what he described as inaccuracies in an entry about himself.
Reportedly told The Washington Post before launch that he had low expectations, arguing language models aren't sophisticated enough yet and that Grokipedia would contain many errors.
One year later — open only if you can handle spoilers
Grokipedia's reach came less from human readers than from other AI systems: by January 2026 ChatGPT's GPT-5.2 and Google's AI Overviews and Gemini were citing it as a source, raising misinformation concerns. A February 2026 Tow Center analysis found Grok itself was by then suggesting and approving most changes to the site.
Filed
Compiled and edited by Adam Allcock.
The Weekly Replay · free by email
This week, one year ago — every Sunday.
One email each Sunday: the week's replayed AI news, with the one-year-later annotations included. Written like it's breaking — dated like it isn't.
Free · double opt-in · unsubscribe anytime · privacy