Google and Microsoft’s AI are also accused of using Grokipedia, not just ChatGPT.

Brief Summary

A recent study has shown that it isn’t just ChatGPT that pulls from the controversial “Grokipedia” encyclopedia created by Elon Musk’s xAI company. Likewise, Google Gemini and Microsoft Copilot have been found to reference it in their answers. The findings were reported on *The Verge*.

1. What is Grokipedia?
- An encyclopedia built by the xAI team (Elon Musk’s company).
- Unlike Wikipedia, it isn’t edited by humans; its content is vetted by its own chat‑bots called Grok.
- Users can flag errors via a pop‑up window, but factual inaccuracies and bias have already been found in articles—e.g., justifying slavery in the United States or softening controversial aspects of Musk’s biography.

2. How often do GPT models use it?
| System | Grokipedia Mentions | Number of Queries |
|--------|---------------------|-------------------|
| ChatGPT | 263 000 | 13.6 million |
| Google Gemini | 8 600 | – |
| Microsoft Copilot | 7 700 | – |
| AI‑reviews in Google (search overviews) | 567 | – |

*In comparison, the English Wikipedia was cited in 2.9 million answers.*

Glen Olsson of Ahrefs noted that since late October 2025 the share of Grokipedia in responses has been rising—from 0.01 % to 0.02 % daily—and by mid‑November it had become noticeably higher.

3. How is the source handled?
- Google services treat Grokipedia as one of many reference materials among a multitude of links.
- ChatGPT, however, often gives it high priority and lists it among the first cited sources, especially for niche or highly specific facts (see Jim Yu from BrightEdge).

4. Why is this important?
Experts warn about the danger of spreading misinformation:

* Grokipedia bypasses traditional editorial verification; mistakes remain unnoticed until a user flags them.
* The presence of biased statements—such as justifying slavery—can mislead readers.

Conclusion: While Grokipedia attracts attention because of its ties to Elon Musk, its use by AI systems poses a risk of disseminating inaccurate and biased information. It is crucial to keep this in mind when selecting sources for generative models.