ChatGPT isn’t the only chatbot pulling answers from Elon Musk’s Grokipedia

ChatGPT uses Grokipedia As a sourceAnd it’s not the only AI tool doing this. Citations to Elon Musk’s AI-generated encyclopedia have started appearing in answers from Google’s AI Overviews, AI Mode, and Gemini as well. Data suggests this is on the rise, raising concerns about accuracy and misinformation as Musk seeks To reshape reality in his image.

Since the Wikipedia clone is distorted It was launched late last OctoberHowever, Grokipedia remains technically a secondary source of information overall. said Glen Allsopp, head of marketing strategy and research at SEO company Ahrefs Edge The company’s testing found that Grokipedia was referenced in more than 263,000 responses to ChatGPT from 13.6 million claims, citing nearly 95,000 individual Grokipedia pages. By comparison, Allsop said that the English-language Wikipedia appeared in 2.9 million responses. “They’re so far away, but it’s still impressive how new they are,” he said.

Based on a dataset tracking billions of citations, a researcher at marketing platform Profound, Sartaj Rajpal, said Grokipedia received about 0.01 to 0.02 percent of all ChatGPT citations daily — a small share but one that has increased steadily since mid-November.

Semrush, which tracks how brands appear in AI answers for Google tools through its AI Visibility Toolkit, found a similar move in Grokipedia’s visibility in AI answers as of December, but noted that it’s still largely a secondary source compared to established reference platforms like Wikipedia.

Grokipedia citations appear on ChatGPT more than any other platform used by analysts Edge I talked to Trace. However, Semrush found a similar spike in Google’s AI products — Gemini, AI Overviews, and AI Mode — in December. Ahrefs’ Allsopp said Grokipedia was mentioned in about 8,600 answers for Gemini, 567 answers for AI Overview, 7,700 answers for Copilot, and two answers for Perplexity, from about 9.5 million, 120 million, 14 million, and 14 million prompts, respectively, with appearances in Gemini and Perplexity down significantly from similar tests the previous month. None of the companies Edge Citations for Anthropic’s Claude are planned to be tracked numerous Nonfiction Reports On social media, the chatbot also cites Grokipedia as a source.

In many cases, AI tools appear to cite Grokipedia to answer specific, vague, or very specific factual questions, as is the case with The Guardian I mentioned Late last week. Analysts agree. said Jim Yu, CEO of analytics company BrightEdge Edge ChatGPT and AI Overviews use Grokipedia for largely “non-sensitive queries” such as searches and encyclopedia definitions, although differences appear in the amount of authority they give it. As for overviews of AI, Yu said Grokipedia doesn’t tend to stand alone, and “usually appears alongside many other sources” as a “supplemental reference rather than a primary source.” When ChatGPT uses Grokipedia as a source, it gives it much more authority, Yu said, “and often displays it as one of the first sources cited for a query.”

Even for relatively mundane uses, experts warn that using Grokipedia as a source risks spreading misinformation and promoting partisan talking points. Unlike Wikipedia, which is edited by humans in a transparent process, Grokipedia is produced by xAI’s Grok chatbot. Grok is perhaps best known for it Nazi collapse, She calls herself Mikahtler, The apotheosis of muskAnd recently, Stripping people digitally onlineincluding minors. When it launched, the bulk of it was Grokipedia articles Direct clones Wikipedia, though many others Reflects racist and transphobic views. For example, articles about Musk downplay his family’s wealth and unsavory elements from their past (Such as neo-Nazi and pro-apartheid views) and incorrectly inserted “gay pornography.” linked Substances that exacerbated the HIV/AIDS epidemic in the 1980s. The article on slavery in the United States still contains a long section on “ideological justifications,” including “the transformation from necessary evil to positive good.” Editing is He is also supervised by Grok and is similarly flawed. Grokipedia is more vulnerable to what is known as “LLM grooming” or data poisoning.

In a comment to Edge“When ChatGPT searches the web, it aims to leverage a wide range of publicly available sources and perspectives relevant to the user’s question,” said OpenAI spokesperson Chaouki Amdo. Amdo also said that users can see the sources and judge them for themselves: “We apply security filters to reduce the risk of links showing up with high-risk damage, and ChatGPT clearly shows which sources informed the response through citations, allowing users to explore and evaluate the reliability of sources directly.”

Perplexity spokesman Bijoli Shah did not comment on the risks of co-opting LLM or citing AI-generated material such as Grokipedia, but said the company’s “central advantage in search is accuracy,” which it is “relentlessly focused on.” Anthropic declined to respond officially. XAI did not return EdgeRequest for comment. Google declined to comment.

The point here is that Grokipedia can never be relied upon as a reliable source, no matter how rare it is and despite Musk taking a hard line. A baseless victory lap About the encyclopedia’s alleged massive success in Google search results. It is a system created by artificial intelligence, lacks human oversight, and often relies on opaque, difficult-to-verify material such as personal websites and blog posts, and questionable, Potentially circular,sources. There is a real risk of reinforcing various biases, errors or framing of issues if you cite something like Grokipedia, said Taha Yasri, head of technology and society at Trinity College Dublin, adding that “it’s easy to confuse fluency with reliability.”

“Grokipedia seems like a costume for credibility,” said Lee McKenzie, director of online impressions at Semrush. “This idea might work within its own bubble, but the idea of ​​Google or OpenAI treating something like Grokipedia as a serious virtual reference layer at scale is bleak.”