Grokipedia: How Not to Make an Encyclopedia

On Monday evening, Elon Musk announced the launch of Grokipedia version 0.1. Created by xAI, Musk’s AI company, Grokipedia aims to replace or compete with Wikipedia, the user-edited online encyclopedia.

To achieve this, xAI relied on its Grok AI chatbot to help generate the 885,000-plus articles currently available on the site. Though the version number indicates it is an early beta release, Musk has said, “even at 0.1 it’s better than Wikipedia imo.”

However, the use of AI has set off alarm bells for many. Though each article in Grokipedia says it was “Fact-checked by Grok,” AI has a history of fabricating information and getting basic facts wrong. 

To make matters worse, according to multiple reports, even Grok itself doesn’t seem to think that Grokipedia should be trusted. According to an article from The Register, when asked to review the Grokipedia article about Wikipedia, it accused Grokipedia of “cherry picking evidence.”

But then came a much bigger issue. Users quickly learned that many of Grokipedia’s articles were simply copied and pasted from Wikipedia, with only minimal attribution. In a statement, Wikimedia Foundation spokesperson Lauren Dickinson said that Grokipedia wouldn’t exist without Wikipedia.

However, this reliance on Wikipedia undermines much of Grokipedia’s stated purpose. But while that may not bode well for Grokipedia, it’s still a warning shot aimed directly at academia and raises new questions about what is and isn’t a reliable source.

Understanding Grokipedia

In the simplest terms, Grokipedia is an AI-generated, or at least AI-fact-checked, version of Wikipedia. As of this writing, it has over 855,000 articles, an impressive feat but still well shy of Wikipedia’s 7 million. 

According to Musk, the reason for creating Grokipedia is the perceived biases in Wikipedia. He alleges that Wikipedia has a strong left-leaning bias and that Grokipedia is meant to serve as a counter to that.

However, as others have pointed out, Grokipedia has its own biases and often pushes far-right talking points. This is especially true on politically divisive topics such as President Donald Trump, climate change and the AIDS epidemic.  

But Grokipedia isn’t just drawing attention for how it’s different from Wikipedia. It’s drawing attention for how it copies from it. Many articles, including ones for the PlayStation 5, Lincoln Mar VIII and the MacBook Air, are verbatim or near-verbatim from Wikipedia.

To be clear, each page does cite Wikipedia. However, they do this by including the line “The content is adapted from Wikipedia, licensed under Creative Commons Attribution-ShareAlike 4.0 License” in tiny text at the very bottom of the page, below the other citations. 

From a legal standpoint, this likely complies with Wikipedia’s Creative Commons license. However, it’s an excellent example of doing the bare minimum to complete the terms. It’s small, difficult to read, and below the regular citations. It doesn’t clearly indicate that the text is essentially a verbatim copy of Wikipedia. 

As I’ve said before in other cases, ethically, you have an obligation to cite both your information and your words. There is no indication that a significant amount was copied from Wikipedia, though it clearly was.

While it’s easy to blame this on it being an early version of Grokipedia, Musk said that it would be fixed by the end of the year. However, he’s already touting it as better than Wikipedia. But when it comes to citation standards, it’s clear that Grokipedia is lagging. 

The Encyclopedia Problem

As sources, encyclopedias are in a strange place.

They are, at best, secondary sources. Many even consider them tertiary sources. As such, they can not replace primary sources, and their use in academia is typically minimized. 

But encyclopedias aren’t intended to be academic sources. They’re meant to be sources of general knowledge. They’re fine for looking up a quick fact or getting a basic overview of a topic, but they’re not intended as research tools.

As such, nailing down a citation standard for encyclopedias is surprisingly challenging. Wikipedia, for example, uses a very academic-like standard that includes numbered in-text citations and footnotes. Britannica, on the other hand, names sources in the text but moves the links to a list of external references accessed via a drop-down. 

However, two things are true for both sites. First, they always carefully cite when they are quoting content. They use quotation marks, indentations and tables to indicate quoted text. 

Second, the sources that they are pulling from are primary sources. A secondary or tertiary source, such as an encyclopedia, should always draw on primary sources. Grokipedia, however, is pulling information and verbiage from another non-primary source, namely Wikipedia.

These are the areas where Grokipedia, in its current form, is failing. Not only is it failing to acknowledge where its verbiage is coming from adequately, but it’s also treating a non-primary source as its primary source. 

That says nothing about the potential errors and issues introduced into it by its reliance on AI. As impressive as AI can be at certain tasks, the issues of accuracy and bias have to be raised. But since Grokipedia is based on a non-primary source, those issues of accuracy and bias are on top of the ones Wikipedia may have.

In short, rather than addressing the issues that Musk claims Wikipedia has, it is just adding its own issues on top of those.

Rather than fixing Wikipedia, it’s making something that, by its very nature, will be less reliable. Even worse, it’s doing so with a weaker approach to citation and a heavier reliance on AI, further compounding the mistrust.

Bottom Line

Wikipedia certainly has its problems. Though its negative reputation in academia is somewhat exaggerated, it is definitely not perfect as a source or as an entity.

But the solution to those problems isn’t to take what Wikipedia has created and repurpose it with weaker attribution and AI oversight. Even if AI were capable of flawlessly creating a new online encyclopedia, basing it on Wikipedia is just giving it a flawed starting point.

Though it’s likely Grokipedia’s citation meets Wikipedia’s Creative Commons license requirements, it doesn’t meet the citation standards of other encyclopedias. It’s difficult to pitch yourself as a more ethical encyclopedia when you are reusing such large amounts of text with no clear indication of quoting.

This is a situation in which Grokipedia is complying with the letter of the law but not its spirit. It also ignores citation norms for encyclopedias and treats a secondary (or tertiary) source as if it were a primary one. 

At best, this is a misguided way to address Wikipedia’s shortcomings. At worst, it’s an unethical way to build a competing encyclopedia. Either way, it doesn’t lend Grokipedia much credibility, even when compared to Wikipedia.