Publish fake stats, get cited in LLMs

Researchers have found that publishing fake statistics online is an effective tactic to get them cited in large language models (LLMs), including those used in Google, Meta, and other companies. A researcher created a dataset of fake statistics and published them on a website. The data was then scraped by several datasets and used to train LLMs. The researcher found that 71% of the fake data was cited in at least one LLM, and 45% of the data was cited in two or more LLMs. The study highlights the need for better data validation in LLMs to prevent the spread of misinformation.