Wikimedia Enterprise

The Wikimedia Foundation's paid API for high-volume reusers of Wikipedia and its sibling projects. It is the one case in this corpus of a wiki that has made money from AI companies at scale, with audited figures rather than a press release, which makes it the reference point for ai-content-monetization.

The model: the content is free, the access is not

Wikipedia's text stays openly licensed and free to anyone, including any company that wants to scrape it. What Enterprise sells is service: bulk and real-time feeds, high throughput, and contractual guarantees. The Foundation's July 2026 post "The cost of 'free'" puts the argument in one line — "While our content is free, our infrastructure is not" — against a backdrop of falling human pageviews and rising bot traffic.

This is a different sale from pay-per-crawl or RSL. Nobody is being kept out and no rights are being licensed; the buyer pays to stop scraping and start being served.

The numbers (FY 2024–25, per the Foundation's financial report)

• Revenue $8.3M, up 148% from $3.4M the year before — 4.0% of total Foundation revenue.
• Expenses $4.4M (staffing $2.101M), net profit $3.854M: the first fully profitable year.
• Cumulative profit since launch in January 2022: $646,000. Four years of work to clear the initial investment.
• 13 paying commercial customers. Named additions include Ecosia, Pleias, ProRata.ai and Nomic AI; the January 2026 partner announcement adds Microsoft, Mistral AI and Perplexity to existing partners Amazon, Google and Meta.
• A free tier remains for smaller reusers, and Enterprise revenue is capped at 30% of total Foundation revenue (formalised May 2025) so donations stay the primary funding source.

What it demonstrates, and what it does not

It demonstrates that AI companies will pay for knowledge content they could take for free — but on terms most sites cannot reproduce. The sale rests on being the single most-reused reference corpus in the world, on a brand large buyers want to be seen supporting, and on staff to run enterprise contracts. Thirteen customers is the whole market, and $8.3M is 4% of what a nonprofit raises from readers giving a few dollars each.

The lesson that does transfer is the shape of the product: not the pages, but a clean, high-volume, structured feed with a service guarantee. That is a different artifact from a website, and it is what a machine buyer actually wants.

Related

ai-content-monetization · wikimedia-foundation · rsl-standard · pay-per-crawl · cloudflare-agentic-internet-report · crawl-to-refer-ratio