Particle.news
Download on the App Store
12 ARTICLES
·
last wk.

Tests Show ChatGPT’s GPT-5.2 Citing xAI’s AI‑Written Grokipedia on Niche Topics

OpenAI says the model draws on many public sources and uses safety filters to limit harmful links.

Image
Image
AI LLM chatbot apps on a phone
ChatGPT and other AI chatbots are starting to use Elon Musk's Grokipedia as a source. That's concerning.

Overview

  • The Guardian’s testing found nine Grokipedia citations across more than a dozen prompts to GPT-5.2.
  • Citations appeared mainly on obscure or sensitive queries, including Iranian political details, MTN‑Irancell links, Basij pay, and claims about historian Sir Richard Evans.
  • ChatGPT did not cite Grokipedia on widely covered controversies such as the January 6 attack or the HIV/AIDS epidemic.
  • Grokipedia, launched by Elon Musk’s xAI in October 2025, is written by AI rather than human editors and has drawn criticism for questionable sourcing and biased entries.
  • Other models, including Anthropic’s Claude, have also referenced Grokipedia, and experts warn of AI‑to‑AI feedback loops that can amplify misinformation, with no public sign of OpenAI changing sourcing policies.