Wikipedia banned large language models after AI content spread
By March 2026, Wikipedia had barred large language models from article writing and limited them to translation and light proofreading.
Wikipedia spent years trying to contain AI before drawing a hard line.
The Wikimedia Foundation and volunteer editors first hoped machine tools might handle routine work, like formatting, citations, and missing links. That bet did not hold.
By March 2026, writing with large language models was banned outright.
A Princeton University study published in October 2024 found that about 5% of nearly 3,000 newly created English Wikipedia articles from August 2024 contained substantial AI-generated text.
The signs were often easy to spot. Some pages carried the style of ChatGPT, with em dashes, curly quotes, and even prompt leftovers such as “Here is your Wikipedia article,”.
Editors tried to clean up the mess. They removed weak filler, including praise for villages with fertile farmland and scenic views, and they cut fake sources.
By March 2026, the project had gone further and banned writing with large language models altogether. Their use was limited to translation and minor proofreading.
That shift reflected a bigger fear inside the Wikipedia world. If AI keeps replacing human contributors, the pool of experienced editors could shrink, and the site could end up with more weak pages and fewer people able to fix them.
The source also argues that a machine-edited encyclopedia would not just be inaccurate. It would risk losing the human context, judgment, and energy that make Wikipedia work at all. What happens to an encyclopedia when the cleanup bot starts writing the article too?
The piece places Wikipedia inside a longer history of dreams about total knowledge, from H. G. Wells to the early ideals behind Wikipedia itself. But the current problem is practical, not abstract: AI is already inside the editing process, and Wikipedia has responded by pushing it back out.
The excerpt does not show how well the March 2026 ban has worked in practice, or how often editors still use AI for the allowed tasks.
Source: WIRED: Artificial Intelligence