Grokipedia, Elon Musk’s AI-written alternative to Wikipedia, is filled with factual mistakes, unnecessary detail and awkward prose, according to Bill Adair, a journalist and Duke University professor who examined the site’s entries about him.
One Grokipedia article incorrectly merged Adair’s biography with that of Marion Danne “Bill” Adair, a minor league baseball player and manager who died in 2002. The page claimed that the journalist had played baseball for 19 years, managed the Chicago White Sox and attended Arizona State University. In reality, Adair says he never played minor league baseball and is 5 feet 8 inches tall, not 6 feet.
Grokipedia also has a separate article about the journalist. Adair counted 32 errors in that entry, including invented details, outdated information and claims that misunderstood his work. The article incorrectly identified him as the director of Duke’s journalism program, a position he left five years earlier, and credited him with breaking the story about George W. Bush’s drunken-driving conviction in 2000.
The site’s “fact-checked by Grok” label did not prevent the mistakes. Adair said many of the problems resulted from the system’s failure to understand the material it had gathered, rather than simply inventing facts. He described Grok as combining details from numerous web pages without reliably judging their importance or accuracy.
The writing was another major weakness. The journalist’s Grokipedia profile ran to about 5,600 words, compared with roughly 700 words on Wikipedia. According to Adair, it repeated information about PolitiFact, explained the organization’s rating system multiple times and relied on corporate-sounding language and buzzwords.
Despite those shortcomings, Adair argued that Grokipedia demonstrates several ways AI could assist Wikipedia editors. Its broad searches sometimes uncover information that human contributors have missed. The Grokipedia article included substantial material about Adair’s research into political lying and his work with fact-checkers internationally, topics absent from his Wikipedia entry.
Grokipedia also included a section on criticisms and controversies surrounding PolitiFact, including accusations that the organization is biased against Republicans and Adair’s response. Although the presentation could be overly enthusiastic, Adair said this kind of context and balance could be useful to Wikipedia, which has faced criticism from Musk and others who view it as too left-leaning.
Musk launched Grokipedia as an alternative to Wikipedia, which he has called “Wokepedia.” Grok reportedly copied hundreds of thousands of Wikipedia articles under the encyclopedia’s Creative Commons license before adding material from elsewhere online. Musk said the system was designed to review popular Wikipedia articles and add, modify or delete content.
The project has yet to match its ambitions. Grokipedia’s articles are often outdated and have been criticized for errors, conspiracy theories, low-credibility sources and a rightward political slant. Its process is also less transparent than Wikipedia’s: while readers can request edits and see whether those requests are approved, Grokipedia does not provide the same level of information about editors and revisions.
The site reportedly stopped updating for a period last spring before resuming in late September. Its homepage was later redesigned and frequently featured an article about Musk. Grokipedia also includes features such as randomly displayed recent edits and AI-generated audio readings in several voices.
Wikipedia has prohibited content generated by large language models, but AI-written material has nevertheless created additional work for its editors. The site’s new-pages patrol must identify and remove low-quality submissions, while the Wikimedia Foundation has warned that AI-generated answers could reduce traffic to Wikipedia and threaten donations.
Adair’s assessment is that AI is not ready to replace Wikipedia’s volunteer editors. Grokipedia’s errors, weak prose and inability to distinguish significant information from trivial details underline the importance of human judgment. At the same time, its ability to search widely, collect overlooked material and introduce additional context suggests that AI could become a useful research and writing aid if editors remain responsible for evaluating and curating the results.
Sources: Nieman Lab.