Distrust in Algorithms: Wikipedia Permanently Bans AI-Generated Text

The Wikipedia community has finalized an update to its rules prohibiting the use of neural networks (ChatGPT, Gemini, DeepSeek, and others) for creating or significantly rewriting encyclopedia texts. Editors are now only allowed to use AI for narrow auxiliary purposes: checking spelling, basic proofreading of their own notes, or machine translation from other language versions.

The decision is the result of lengthy discussions within the community. The primary reason is the incompatibility of how modern LLMs (Large Language Models) operate with the encyclopedia’s fundamental principles. Wikipedia requires that information be based on verifiable, authoritative sources and maintain a neutral point of view, but neural networks are known for their tendency to "hallucinate." As noted in the updated guidelines, using AI to generate text systematically leads to violations of these core principles. Even if an editor polishes AI-generated text, there’s a high risk that the model will subtly slip in and distort the meaning by adding information not found in the cited sources.

Special emphasis is placed on the fact that editors should not rely solely on stylistic cues ("it sounds like a robot") when trying to flag a colleague. Decisions regarding violations must be based on analyzing the factual accuracy of the text and its alignment with the sources. For now, the restriction only applies to the English-language segment, which is the largest and most influential. Other language versions, including the Spanish one—which introduced a ban earlier—may adhere to their own, sometimes stricter, policies.

This change is the latest step in the community’s efforts to combat an influx of low-quality content. Over the past few months, editors have been actively creating special projects (such as WikiProject AI Cleanup) to scrub articles that were generated by neural networks in the early stages of their proliferation.

For Wikipedia, which recently celebrated its 25th anniversary, this decision appears to be an attempt to preserve reader trust in an era of rapidly spreading synthetic text. The platform, which garners billions of views monthly, is choosing to side with verifiable human labor over the speed of generating content that could contain fabricated facts.

What do you think—will the Wikipedia community be able to effectively track down violators, or will moderators now have to spend more time on investigations than on writing the articles themselves?