Wikipedia and AI bots: It hasn’t blocked artificial intelligence, but it’s limiting abuse

Wikipedia has not banned artificial intelligence or blocked major tech companies. The Wikimedia Foundation is taking a more precise approach: it is restricting automated requests that violate its rules and overload its servers. At the same time, it offers data through paid partnerships and a dataset designed for training. This article explains what really happened, why it’s important for the open web, and what this means for copyright.

What’s Verified and What Isn’t

• Rule changes at the end of October 2026: I found no evidence of a new blanket change to robots.txt or API rules from late October. The foundation announced significant steps earlier this year (see below).
• Blanket blocking of large companies: this has not been confirmed. The foundation only raised the suspicion that specific problematic agents might have been operated by a single company. I am not aware of any blanket blocking of multiple large companies.
• A ban on “AI” as a whole: The foundation emphasizes that human access should remain unrestricted and that content should be freely available and freely licensed. The target is abusive bots, not artificial intelligence as such.

What the foundation actually did

• March and April 2026: The foundation announced that it was blocking or throttling approximately one-quarter to one-third of automated requests that violated the terms of use. It implemented global limits on the number of API requests in two phases. Various sources cite a range of 25 to 30 percent, as the measurements differ.
• A common reason: bots are more costly for the infrastructure than humans. They access many low-traffic pages at once and cannot be easily cached locally. According to the foundation, this has led to an increase in load since the beginning of 2024. For multimedia content, bandwidth consumption rose by 50%, and bots accounted for approximately two-thirds of the most resource-intensive traffic, even though they accounted for only one-third of page views.
• October 2026: On October 5, the foundation reported that agents—which it believed were likely operated by OpenAI—had edited certain wikis without the approval required by the rules. At the same time, they unsuccessfully attempted to exploit a public annotation tool and sent millions of requests to the public API. The Foundation also linked them to the outage. OpenAI stated that it is cooperating with the Foundation on the analysis. This is a suspicion on the part of the Foundation, not a judicial finding.
• Voluntary approach: The Foundation offers a dataset intended for AI training and formalizes paid partnerships with several technology companies to provide them with simplified access to the data.
• Content: English Wikipedia prohibits articles written by artificial intelligence. This is an independent decision by the community.

Why robots.txt Isn't Enough

The robots.txt file is a convention, not a lock. There are bots that ignore it, and some site operators argue that the rules may not apply to agents whose activity is initiated by a user. Wikipedia therefore relies on technical restrictions (request limits), bot identification rules, and approval. The rules for editing bots are enforced through accounts and community approval, not through the file.

A Question of Fairness

The crux of the dispute is not whether AI is allowed to read Wikipedia. The content is freely licensed, and the license permits distribution and modification under certain conditions (attribution, same license). Two issues are in dispute:

• Costs. A foundation funded by donations and volunteers bears the operating costs incurred by commercial systems. It argues that those who profit from the data should contribute.
• Attribution. The license requires attribution to the authors. It is unclear how this applies to the outputs of language models. This legal issue remains unresolved in many countries, and this article does not address it.

Copyright and AI training are the subject of legal disputes and regulation. The outcome varies by country, and no general rule can be derived from it.

What does this mean for the average reader?

• Wikipedia remains freely accessible. No digital wall has been erected for people.
• The costs of running open projects are rising, and this trend also affects other platforms and smaller websites.
• If you use an AI tool that crawls the web, its operator should respect the access rules. The Foundation requires that bots be clearly identified.

What to Take Away From This

The headline “Wikipedia Has Blocked Artificial Intelligence” is an oversimplification. The real story is about infrastructure, costs, and the question of who pays for the open knowledge that fuels today’s commercial models. The foundation’s response is not a ban, but a combination of restrictions on abusive bots and paid or negotiated pathways for those who need data at scale.