Publish fake stats, get cited in LLMs

A researcher created a website with 50,000 fabricated statistics to test how large language models like ChatGPT discover and cite information, finding that the fake data was picked up by over 330 referring domains including major companies and appeared in academic papers and investment research. The experiment, which cost only $50 to set up, reveals a significant vulnerability in how AI systems validate and propagate information, as ChatGPT continues to direct thousands of sessions to the site despite Google deindexing it.