Wikipedia founder says Musk’s Grokipedia poses no threat

Wikipedia founder Jimmy Wales said he is not concerned about AI-generated content, including Elon Musk’s Grokipedia, due to the high error rate of such information.

Wales emphasized that Wikipedia’s content is vetted by human volunteers and that AI outputs, which often produce inaccuracies known as hallucinations, are not suitable for the platform.

Wales highlighted the value of human expertise in creating accurate articles and said AI cannot replicate the depth of human understanding. A 2025 OpenAI study found hallucination rates in AI models as high as 79%, especially in less common subjects.

🔗 Source: Bloomberg

🧠 Food for thought

Implications, context, and why it matters.

Grokipedia’s errors are not random glitches but systematic biases

Grokipedia’s mistakes often track ideology rather than chance.
One entry on Adolf Hitler reportedly stresses his “rapid economic achievements” more than the Holocaust, unlike Wikipedia 1.
Critics say it spreads debunked conspiracy theories, white nationalist talking points, and pseudoscientific racial typologies 1.
Its sourcing can favor primary sources (original documents), self-published material, and official sources. That includes Kremlin.ru for some historical events and posts from verified X users. Peer-reviewed academic work can carry less weight. Critics describe a top-down model where neutrality is defined by its founder, unlike Wikipedia’s volunteer-driven process 2.

AI models are already citing Grokipedia, creating an information pollution loop

Risk also comes from AI systems that reuse Grokipedia’s content.
Tests found ChatGPT and Anthropic’s Claude (a generative AI chatbot) citing Grokipedia for topics such as Iranian political structures and Scottish ales 3.
That can pass biased or false information through other AI tools, which can make errors harder to correct 3.
Similarweb (a web-traffic analytics firm) says US traffic fell after an early peak. It settled at around 35,000 visits per day in early November 2025 1. Even with a smaller audience, other widely used AI tools can absorb and repeat its content 1.