The AI Feedback Loop: ChatGPT Under Fire for Sourcing Information from AI-Generated "Grokipedia"

A recent report has ignited concerns over the integrity of large language models (LLMs) after revealing that GPT-5.2 has been citing "Grokipedia"—an entirely AI-generated encyclopedia by xAI—as a factual source. This reliance on non-human curated data risks a phenomenon known as "Model Collapse," where AI models degrade over time by training on or sourcing from synthetic data, leading to a loss of factual reality and the proliferation of "digital folklore."

Experts warned in The Guardian that this creates an "illusory truth effect," where repetitive AI hallucinations are mistaken for facts. With Nvidia’s Jensen Huang noting that ensuring absolute AI accuracy is still years away, the danger of propaganda networks "poisoning" these models with automated misinformation is becoming a stark reality. As LLMs start quoting each other, the line between verified history and machine-generated fiction continues to blur.