Quick Answer: Wikipedia AI deals are finally happening. Amazon, Meta, Microsoft, Mistral AI, and Perplexity just signed paid data agreements with the nonprofit after 25 years of free access. Wikipedia’s founder says these companies have been “hammering their servers” and it’s time they chip in.
So Wikipedia turned 25 this month and I now feel old because I remember its inception. A quarter century. Free the whole time.
Anyone could grab its content, republish it, build entire businesses on top of it, and now the world’s biggest encyclopedia is finally passing the hat around to AI companies. I’m surprised it took this long.
Wikipedia AI Deals with Amazon, Meta, Microsoft and More
Amazon, Meta, Microsoft, Mistral AI, and Perplexity signed paid data access agreements with the Wikimedia Foundation through Wikimedia Enterprise, joining Google which has been paying since 2022.
CNBC broke the story on January 15th. Amazon, Meta, Microsoft, Mistral AI, and Perplexity are all signing up. All agreeing to pay for data access through something called Wikimedia Enterprise instead of just scraping it like they used to.
The timing feels deliberate. January 15, 2026 was Wikipedia’s 25th birthday. What better gift to yourself than finally getting paid?
Google actually jumped on board back in 2022. I didn’t know that until I started digging into this. The rest of these Wikipedia AI deals were apparently cooking for a year before anyone said anything publicly.
Why Wikipedia Started Charging AI Companies
Wikipedia founder Jimmy Wales told Euronews that AI companies have been “absolutely hammering” Wikipedia’s servers, and the donation-funded nonprofit with 65 million articles and 15 billion monthly pageviews can no longer absorb the cost of training the world’s AI models for free.
Jimmy Wales didn’t sugarcoat it. He told Euronews that AI companies have been “absolutely hammering our servers” and that Wikipedia has been encouraging them to sign up for enterprise products instead.
Then he got blunt. “You should probably chip in and pay for your fair share of the cost that you’re putting on us.”
Translation. Stop freeloading.
People forget that Wikipedia runs entirely on donations. No ads. No investors. Just regular folks throwing in a few bucks when that annual banner guilt trips them into it. We’ve all seen it and somehow this nonprofit ended up in the top 10 most visited websites on Earth. The only nonprofit on that list. Wikipedia has a whopping 65 million articles across more than 300 languages. Nearly 15 billion pageviews every single month.
This was fine until AI companies showed up and started vacuuming all that content into their training data.
Wikipedia’s Traffic Problem After AI
Wikipedia’s human traffic dropped about 8% year over year because people now ask ChatGPT, Claude, or Perplexity instead of clicking through to read the source articles that trained those models in the first place.
This part caught me off guard. MediaPost says Wikipedia’s human traffic dropped about 8% compared to the same stretch in 2024.
Eight percent. That matters.
Part of it was their own doing. Wikipedia tweaked some algorithms to block bot traffic and accidentally caught some real humans in the net too. But there’s a bigger issue lurking here. People just don’t visit Wikipedia like they used to because they ask ChatGPT instead.
Think about your own habits for a second. When you ask an AI a random question, where do you think that answer came from? A lot of it traces back to Wikipedia. But you never clicked through. You never saw their donation banner. Wikipedia did all the work, got zero credit, lost the traffic. Pretty raw deal when you step back and look at it.
I catch myself doing this constantly. Ask NotebookLM something, get an answer, move on. Never once thinking about who actually wrote the source material.
What Microsoft Says About Wikipedia AI Deals
Microsoft corporate VP Tim Frank called trustworthy information “at the heart of how we think about the future of AI,” acknowledging that Wikipedia is some of the cleanest, most reliable training text on the internet.
TechCrunch got a quote from Tim Frank, a corporate VP at Microsoft. He said access to trustworthy information is “at the heart of how we think about the future of AI at Microsoft.”
Corporate speak aside, the subtext is clear. Microsoft’s AI is only as smart as the stuff it learned from. Wikipedia happens to be some of the cleanest, most reliable text on the entire internet. Paying for proper access just makes business sense.
Worth mentioning that the New York Times is currently suing Perplexity over this exact kind of scraping. Wikipedia going the partnership route instead of the lawsuit route seems way smarter in the long run. Less messy. Less expensive. Everybody walks away with something.
What Wikipedia AI Deals Mean for You
AI companies get cleaner data feeds instead of janky scrapes, Wikipedia gets sustainable funding, your AI responses get slightly more accurate, and the nonprofit model gets a fighting chance in an AI-dominated world.
Alright, so why should you care about any of this?
If you’ve ever touched ChatGPT, Claude, Perplexity, or basically any AI assistant, you’ve indirectly used Wikipedia. Those models learned from it. Every time they spit out an answer, there’s a decent chance Wikipedia’s volunteer editors did some of the heavy lifting years ago.
These Wikipedia AI deals shake out a few ways.
- AI companies get cleaner, more organized data instead of janky scrapes
- Wikipedia actually gets money to keep the lights on
- Your AI responses get slightly more accurate because of better data access
- The nonprofit model gets a fighting chance in an AI dominated world
You probably won’t notice anything different next time you ask Claude a question. But somewhere in the background, the whole economics of how AI accesses information is quietly shifting.
Wikipedia Gets New Leadership
The Wikimedia Foundation named former US Ambassador to Chile Bernadette Meehan as its new CEO, a diplomatic hire that makes sense given Wikipedia’s biggest challenges now involve navigating partnerships, misinformation, and coexistence with AI companies.
Oh, and one more thing. The Wikimedia Foundation is also swapping CEOs. WinBuzzer reports that Bernadette Meehan takes over January 20th. She used to be the US ambassador to Chile.
An ambassador running an encyclopedia. Sounds weird at first. But when you think about it, Wikipedia’s biggest headaches right now are basically diplomatic. How do you stay open while also staying funded? How do you fight misinformation without becoming a gatekeeper? How do you coexist with AI companies that are simultaneously your biggest users and your biggest threat?
Maybe a diplomat is exactly what they need.
Wikipedia AI Deals FAQ
Is Wikipedia still free to use?
For you and me? Absolutely. Nothing changes there. These deals only hit companies scraping data to train AI. Regular folks can still read all 65 million articles without paying a dime. Same as always.
Which AI companies are paying Wikipedia?
Amazon, Meta, Microsoft, Mistral AI, and Perplexity just signed on. Google has been paying since 2022. All through this Wikimedia Enterprise thing.
Does this mean ChatGPT and Claude use Wikipedia data?
Yep. Pretty much every major AI model trained on Wikipedia to some degree. It’s one of the biggest collections of quality text that exists online.
How much are AI companies paying Wikipedia?
Nobody is saying. The actual dollar amounts haven’t been made public. Wikimedia Enterprise basically offers a clean data feed as an alternative to scraping, but the price tag remains a mystery.
The Bottom Line on Wikipedia AI Deals
After 25 years of giving everything away for free, Wikipedia is finally getting paid by the AI companies that built their products partly on its volunteer-written content, setting a precedent for how AI companies should treat the sources they depend on.
For 25 years, Wikipedia gave everything away. Now the AI companies that got rich partly because of that generosity are finally writing checks.
Feels like a small bit of justice. Wikipedia can keep running. Keep improving. And maybe this sets a precedent for how AI companies should treat the sources they depend on.
Next time you ask an AI to summarize something, remember that volunteers probably wrote the original text it’s pulling from. At least now those volunteers have some corporate money backing them up.
Related reading. Get AI to Explain Complex Topics Simply
Leave a Reply