AI Takes Traffic Away from Wikipedia: Microsoft, Google and Amazon Will Pay for ‘enterprise’ Data Access

On the occasion of Wikipedia’s twenty-fifth anniversary, the Wikimedia Foundation announced a pivotal strategic shift for the survival of the world’s most famous free encyclopedia.

Facing the growing influence of artificial intelligence, which heavily uses the site’s content to train its models, the organization has formalized a series of commercial agreements with the tech giants.

Microsoft, Meta, Amazon, together with emerging realities such as Perplexity and the French Mistral AI, joined Google in funding access to Wikimedia projects, recognizing the economic value of the data that powers their systems.

Wikimedia Foundation, a new economic balance for the AI era

The decision stems from a pragmatic and urgent necessity. Over the past year, the foundation has raised the alarm about a decline in direct traffic to the site, a phenomenon largely attributable to the spread of chatbots and large language models (LLMs).

These tools, capable of providing immediate and comprehensive summaries, discourage users from clicking on the original links, undermining the sustainability model of Wikipedia, which relies on visitors’ donations.

At the same time, the activity of “scraping”, i.e., the automated extraction of data by tech companies to train their AIs, has led to a dramatic increase in server costs for the nonprofit organization.

Lane Becker, president of Wikimedia Enterprise, explained how the situation required structural intervention. The goal was to move these large companies from the free, public-facing platform to a commercial solution that could offset infrastructure costs and ensure the encyclopedia’s operational continuity.

Wikimedia Enterprise: tailored data for Big Tech

The vehicle for this transition is Wikimedia Enterprise, an initiative launched in 2021 that offers a premium service to large companies. Unlike public access, this solution provides high-speed APIs and structured data sets specifically for commercial use and training algorithms.

Becker clarified that the program is not limited to selling access, but envisions active collaboration: the foundation welcomes specific functionality requests from companies and works to organize the data in ways that support their technical needs.

Although Google was already a well-known partner, today’s announcement reveals that Microsoft, Perplexity and Mistral AI joined the program over the past year. Moreover, it is disclosed for the first time that giants such as Meta and Amazon are among the ‘existing’ partners, formalizing a relationship that until now had remained out of the spotlight.

Funds raised through these partnerships are reinvested directly into the nonprofit’s projects, helping to establish a more solid business model and less reliance on donations from individuals.

A Necessary Symbiosis

The logic behind these agreements is mutual dependence. As emphasized by Wikimedia’s leadership, it is in the AI companies’ own interest to support Wikipedia’s longevity.

The foundation’s projects, which also include Wikivoyage, Wikibooks and Wikiquote, indeed represent a large portion of training data for almost all modern AI systems. Without a steady flow of up-to-date, verified, and human-curated information, the quality of the responses provided by chatbots would rapidly degrade.

Achieving a new ‘sustainable balance‘, according to Becker, is therefore critical not only for Wikipedia’s existence, but for the very survival and effectiveness of the services offered by the tech giants.

This agreement thus marks a historic milestone: the official recognition that, in the era of synthetic information, curated and freely available human knowledge has a cost that the world’s wealthiest companies must contribute to sustaining.