Elon Musk’s Grokipedia, the artificial intelligence (AI) encyclopedia, uses right-leaning sources when discussing sensitive topics, according to a new study.
The study, conducted by researchers at Trinity College Dublin and Technological University Dublin, examined nearly 18,000 of the most-edited English-language pages on Wikipedia and compared them with corresponding entries on Grokipedia, the AI-written alternative launched by Elon Musk in October last year.
Grokipedia articles about religion, history, literature and art referenced more right-leaning outlets than their Wikipedia counterparts, the study found, though overall, the Grokipedia and Wikipedia articles showed similar political leanings.
Two-thirds of the Grokipedia articles analysed were heavily rewritten and relied on fewer sources than their Wikipedia equivalents, the study found.
The report comes after a pre-print study from January found similar concerns about Grokipedia, where many articles did show a left-leaning bias overall, but some articles on controversial topics could prioritise right-leaning content.
Meanwhile, the European Commission launched an investigation in January into Musk’s xAI under the Digital Services Act (DSA) to see whether it disseminated illegal content in the EU, such as manipulated sexually explicit images.
The findings highlight how generative AI systems may reshape public knowledge in ways that are less visible than traditional editorial processes, the findings claim.
“Unlike Wikipedia, where biases are visible and contested through human editing, AI-generated systems operate largely opaquely,” said Saeedeh Mohammadi, lead author of the study. “This means shifts in perspective or sourcing may occur without clear accountability or editorial oversight.”
The authors also warn that the rapid expansion of AI-generated knowledge systems raises wider governance questions.
They point to risks similar to those already observed on social media platforms, where limited editorial oversight has contributed to the spread of misinformation with real-world consequences for elections, public health and social stability.
“Our information landscape is changing rapidly,” said Taha Yasseri, professor at Trinity College Dublin. “We are witnessing the large-scale, black-box regeneration of information by large language models that remain largely closed to public scrutiny.”
