Policy

Wikipedia Bars AI From Direct Editing Over Hallucination Risks

Cofounder Jimmy Wales says fabricated output remains a critical problem, even as AI bots drive traffic growth on the platform.

Omega Editorial· June 22, 2026· 2 min read

Wikipedia will not permit artificial intelligence systems to edit articles directly on its platform, cofounder Jimmy Wales confirmed this week, citing ongoing concerns about AI-generated fabrications that undermine the encyclopedia's reliability standards.

Speaking at a climate action event in London, Wales told AFP that AI "hallucinations"—instances where systems confidently present false information—remain "very, very bad" despite improvements in newer models. The risk is too high for Wikipedia's editorial integrity, which depends on accuracy verified by human contributors.

AI as Assistant, Not Editor

While direct editing is off the table, Wales suggested AI agents could serve a supporting role by flagging niche news stories that Wikipedia's millions of volunteer editors might otherwise overlook. This approach would keep humans in control of final editorial decisions while leveraging AI's ability to scan large volumes of information.

The stance reflects a broader tension as AI companies increasingly rely on Wikipedia's content to train models and answer user queries. That dependency has reshaped traffic patterns to the site.

Traffic Shifts and Server Costs

Wikipedia has experienced an eight percent decline in human traffic even as visits from AI bots have grown substantially. Wales characterized the drop as "meaningful" but stopped short of calling it catastrophic for the platform, which ranks among the world's 10 most-visited websites.

Because Wikipedia operates on user donations rather than advertising revenue, the traffic decline doesn't directly threaten its business model. However, the surge in AI bot requests creates real infrastructure costs. "Hammering us with millions of requests costs real money" in server expenses, Wales noted.

He called on AI companies to "pay their fair share" for the value they extract from Wikipedia's content. The Wikimedia Foundation has already secured agreements with several major tech companies, Wales said, and has begun blocking bots from organizations that refuse to cooperate.

Why it matters

Wikipedia's decision highlights a critical challenge for the AI industry: even as large language models improve, their tendency to fabricate information remains incompatible with systems requiring factual accuracy. The encyclopedia's hybrid approach—using AI for discovery while reserving editorial control for humans—may become a template for other knowledge platforms navigating similar reliability concerns. Meanwhile, the infrastructure cost question signals growing friction between AI companies that consume open knowledge resources and the organizations that maintain them.

These details were first reported by AFP during Wales's appearance at a London climate event.

#wikipedia#ai hallucination#content moderation#ai training data#wikimedia foundation#jimmy wales

This is an original analysis by the Omega editorial team. Source reporting: AI Watch.

Want systems like this working for your business?

Book a Call

More in Policy

Policy· 4 min read

House AI Bill Omits Key Provisions, Complicates Senate Talks

A bipartisan House proposal on children's online safety diverges sharply from Senate plans, threatening progress on both AI regulation and tech legislation.

Via AI Watch · Jun 23, 2026
Policy· 4 min read

South Korea's AI chip windfall masks deepening economic divide

Record semiconductor exports drive headline growth, but policymakers warn gains remain concentrated as small businesses and services struggle.

Via AI Watch · Jun 23, 2026
Policy· 3 min read

AI Companies Pour $37M Into Congressional Races Via Super PACs

OpenAI and Anthropic are backing rival networks of political groups that have become top spenders in the 2026 midterms, reflecting a sharp divide over federal AI regulation.

Via AI Watch · Jun 22, 2026