How Artificial Intelligence Could Kill Wikipedia
Editors of the global encyclopedia are fighting AI but are currently losing
In 2018, Google began training AI to create Wikipedia-style articles. By August 2024, approximately 5% of new materials in the English version contained significant fragments written by artificial intelligence. Editors noticed characteristic signs: excessive use of dashes, curly quotation marks, and remnants of prompts such as "Here is your Wikipedia article."
Initially, the community attempted to control AI: removing generic phrases like "the village is renowned for its fertile lands" and fake sources. However, by March 2026, it became clear that machine-generated content too frequently violated the encyclopedia's principles. Editors banned the use of large language models for writing articles, permitting them only for translation and proofreading.
Experts had previously warned that reducing human contribution could lead to editors being replaced by even more AI. One X user bitterly remarked: "I want AI to wipe and wash dishes, not write articles for me." Authors feared that people might end up in the role of service staff for machines.
The community faced the threat of "washing out" talented authors. "The main danger is the reduction of the pool of qualified editors, especially experts," wrote one participant. "Without them, it is impossible to correct the flow of weak, inaccurate, and incomplete articles." A machine-generated encyclopedia risks becoming soulless and devoid of context.
When Google implemented AI overviews in search, they quickly became as important as Wikipedia itself. But if the encyclopedia falls under the influence of artificial intelligence, how can the reliability of answers be verified? Wikipedia's authority as a source of knowledge may suffer.
Former Wikimedia Foundation CEO Katherine Maher warned as early as 2019: "Without people, AI is capable of sowing chaos." She noted that the technology is often presented as a universal solution capable of replacing humans. However, control over machines remains a key problem.
