This article was AI-generated as part of an experimental historical-content project. The date reflects the period being analyzed rather than the date the article was originally written.
The Wikimedia Foundation spent the week of the UN General Assembly in New York making an argument that sounds like a paradox. At the UN’s SDG Media Zone, chief executive Bernadette Meehan put it plainly: “Wikipedia’s content is free. We don’t run ads. We don’t monetize. But our infrastructure is not free.” She tied the cost to scraping and to “the behavior of some of the big technology and frontier AI companies.” The Foundation’s summary of the week, published today, says an AI-dominated internet has made digital public goods like Wikipedia “more vital, but less visible.”
That phrase invites a conclusion I think is wrong, and it is one reputation teams could easily reach.
The myth
It goes like this. Fewer people click through to Wikipedia because search summaries and chatbots answer the question for them. A growing share of the traffic is bots. Human readers were what made a Wikipedia article matter to a company’s reputation. So if fewer humans read it, the article matters less, and the slow work of keeping it accurate can move down the list.
Why it is backwards
The Foundation’s own description of the problem points the other way. The post calls Wikipedia “one of the highest-quality datasets used in training Large Language Models” and says its knowledge forms “the backbone of information on the internet, powering AI chatbots, search engines, voice assistants, and more.” The scraping is heavy because the content is valuable as input.
For a company, that changes where an article’s influence shows up rather than reducing it. A sentence about a company’s history, leadership or controversies used to reach people mainly when they opened the page or saw it in a Knowledge Panel. Now it also reaches them indirectly, absorbed and paraphrased by systems that may not link back. Meehan’s point that unattributed summaries “block pathways” for people to reach and check original sources is a sustainability argument for the Foundation. For a company it is also an accuracy argument. An error that is paraphrased without a source is harder to trace than an error that is quoted with one.
Less visible means less checkable. It does not mean less used.
What the Foundation asked for
The requests are broad. Large commercial reusers should recognize that their products depend on human-created knowledge and support it “through content attribution and financial support.” AI governance should consider inputs as well as outputs, and regulation should avoid harming the public goods the technology relies on. Some of that support already exists: the Foundation’s 25th anniversary announcement in January listed Microsoft, Mistral AI and Perplexity among new Wikimedia Enterprise partners, joining Amazon, Google and Meta.
Nothing in today’s post announces a new rule, a crawler block or a change to how articles are written. It is advocacy, and it is too early to say what it produces.
The practical posture
None of this changes Wikipedia’s norms for companies. Conflict-of-interest guidance still discourages editing your own article directly, and changes still depend on independent, reliable sources raised through the community’s channels. What changes is the reason to care. Read the article about your company and its executives as if it were a source document for every AI system, because in practice it is one. Look for outdated leadership facts, unsourced claims and sections that have drifted out of proportion, and pursue corrections the accepted way.
In 2023 I described the neutral Wikipedia sentence as infrastructure once ChatGPT began citing it. The Foundation is now making a version of that case at the UN, from the other side: infrastructure that more systems rely on and fewer people see. Treating it as optional because it has become harder to see would be exactly the wrong lesson.