As the influence of computer-generated content created by advanced language models such as OpenAI’s GPT continues to grow on the user-generated internet, Wikipedia editors face a new challenge. Beyond their regular task of filtering out poor quality human contributions, they are increasingly required to identify and eliminate content produced by artificial intelligence.

In a recent interview, Ilyas Lebleu, an editor at the community-driven encyclopedia, and a founding member of the “WikiProject AI Cleanup” initiative, shared insights into their work. The team’s mission is to develop effective strategies to identify and manage the influx of machine-generated content. Contrary to what one might expect, artificial intelligence is not of much help in this area.

One of the major issues with AI-generated content is the lack of proper sourcing. The capability of large language models to instantly generate seemingly credible text has resulted in the creation of completely fabricated entries, which are often designed to trick Wikipedia’s human moderators into believing they are legitimate.