Wikipedia and AI: Why Wikipedians Are Fighting Back Against AI-Generated Content

A long time ago, when the AI storm first hit Wikipedia, the Wikimedia Foundation was already preparing. Wikipedians battened down the hatches, confidently predicting that those defenses would eventually break and that they would be swept away.

Large language models could also improve Wikipedia. Tasks such as formatting, attribution, citations and adding missing links might be automated, freeing people to focus on more creative work. But it was not that simple. In 2019, Kathryn Maher, then executive director of the Wikimedia Foundation, wrote a New York Times opinion essay titled “Without Humans, AI Could Wreak Havoc”.

“Too often, artificial intelligence is presented as a one-size-fits-all solution to our problems, a scalable human replacement,” she wrote. The challenge Wikipedia faced in maintaining human control over machines was the same challenge that would soon confront almost everyone.

How AI-generated articles began appearing on Wikipedia

The warning signs were already visible in 2018, when Google trained an AI system to write “Wikipedia-like articles.” By October 2024, a Princeton University study had found that approximately 5 percent of the roughly 3,000 new articles created on English Wikipedia during a single month—August 2024—contained a significant amount of AI-generated content.

The writing often revealed its origins. ChatGPT favored em dashes, parenthetical statements and curly quotation marks. In some cases, AI-generated articles even included fragments of the prompts used to create them, such as “Here’s your Wikipedia article.”

Why Wikipedia restricted large language models

At first, Wikipedians tried to shape and control AI-generated text, making its output “acceptable and constructive” or modifying and removing it when it failed to meet Wikipedia’s standards. Generic passages—such as praise for villages with fertile farmland, livestock and scenic countryside—were rejected. False sources were removed as well.

By March 2026, writing with large language models was completely banned, while AI’s role was limited to translation and small-scale proofreading. The change came after AI-generated content was found to frequently violate Wikipedia’s core principles.

The risk of a downward spiral

A widely read essay titled The Death of Wikipedia explained this defensive posture and predicted a damaging ripple effect. If contributors reduced their participation because AI was infiltrating the platform, the temptation to replace even more human work with AI could create a downward spiral.

A viral post on X captured the frustration: “What I want is for an AI to do my laundry and dishes so I can do my art and writing, not to have an AI do my art and writing so I can do my laundry and dishes.”

The Death of Wikipedia warned against a similar outcome. Instead of handling Wikipedia’s housekeeping, AI could begin creating articles and leave humans responsible for formatting, correction and cleanup. Editors would then become trapped in endless curation projects built around the output of flawed machines rather than human priorities.

Why human editors remain essential

Another Wikipedian wrote, “The potential ‘loss’ of talented writers each year is the biggest threat to Wikipedia.” The writer argued that only an increase in editors—particularly professional editors—could address the “hellish landscape” of weak, inaccurate and incomplete articles holding the project back.

A Wikipedia written and read by machines could become a dilapidated version of itself, stripped not only of context but also of the passion that motivates people to contribute. It could be artless and soulless.

When the Wikimedia Foundation says that “Wikipedians’ care and dedication to trusted encyclopedic knowledge cannot be replaced by AI,” it is not necessarily saying that human contributors cannot be replaced in practice.

What AI-generated errors mean for trust

When Google introduced AI Overviews to search, the feature quickly became central to the search experience, much as Wikipedia once was. Wikipedia still plays an important role as a source of information, even when it appears one step removed through an AI-generated answer.

Anyone who uses AI agents knows that they can produce plausible, almost intuitive answers that pass a superficial review of their sources. They are also often wrong. If Wikipedia itself becomes corrupted by artificial intelligence, how could anyone reliably prove that Wikipedia is wrong?

Could AI create a single machine-edited encyclopedia?

Implicitly, the idea of a single machine-edited encyclopedia ruling them all suggests that such a system would become the final authority. For some people, the appeal lies in the image of a digital King Solomon: cold, rational and seemingly impartial.

There is no reason to assume that such ruthless logic would be harmless. Its aesthetic flaws offer a clue. “Slop” is a useful word for AI products that may appear colorful and substantial while remaining monotonous and lacking nutritional value—or even nutritious flavor.

What Wikipedia means for human flourishing

There is also no reason to assume that AI encyclopedias and existing encyclopedias should be merged. For all its flaws and quirks, Wikipedia remains the closest thing humanity has created to an all-encompassing encyclopedia.

After the iPhone launched in 2007, the widespread adoption of smartphones fulfilled the old, ambitious vision of Wikipedia as cheap, portable, readily available and electronic. It is vast and ubiquitous, and has become something close to “the sum total of human knowledge.”

But what does access to this oracle mean for human flourishing? Have you done Wikipedia?