Wikipedia Tightens Rules To Combat AI-Generated Errors

Wikipedia is warning readers about AI-generated text, speeding page deletions, and banning AI images in a major push to preserve reliability and trust.

The warning label “This article may incorporate text from a large language model” appears on hundreds of Wikipedia articles. The Washington Post reports how volunteer editors work to identify and eliminate AI-generated content containing errors, false citations, and misleading data.

“People really, really trust Wikipedia,” said Lucie-Aimée Kaffee, a policy and AI researcher, as reported by The Post. “And that’s something we shouldn’t erode,” Lucie added.

The accuracy of Wikipedia depends on a worldwide network of volunteers who have fought against spam and vandalism since the platform’s inception. Since 2022, AI text generation has simplified the creation of convincing but flawed Wikipedia articles, said Marshall Miller, director of product at the Wikimedia Foundation. “I like to think of it as like an immune system,” Miller said, as reported by The Post.

A Princeton University study found that about 5% of new English Wikipedia pages in August 2024 contained AI-generated text, as reported by The Post. Some were harmless, but others promoted businesses or political agendas. Examples included a hotel described instead of a village, and a fake Ottoman fortress article that stayed online for nearly a year.

To combat this, editors launched WikiProject AI Cleanup in 2023, creating guides to spot AI writing and listing over 500 suspect articles. In August, Wikipedia implemented a new deletion policy to enable quick page removals, particularly those containing visible AI-produced content, or leftover prompts. Additionally, AI-generated images are banned entirely.

“Oh man, I’ve been finding a LOT of AI slop in the submission queue,” one editor complained in June, as reported by The Post.

While the Wikimedia Foundation has explored AI tools to help editors with tasks like translation, Miller stressed: “The way that Wikipedia remains neutral and reliable … is that all this content passes through the hands of people.” Kaffee added: “Asking, ‘Where do we want AI? [ … ] What kind of rules do we want to set for AI-generated knowledge?’ I think is really important.”