AI IS SO HUNGRY FOR HUMAN WRITING, IT'S CUTTING THE SPINES OFF BOOKS
AI IS BUYING THE BOOKS — AND SOMETIMES DESTROYING THEM
Artificial intelligence was supposed to replace writers. Now AI companies are hunting down human-written books because they need one thing the internet is becoming increasingly contaminated with:
Writing created by actual humans.
Think about how strange this is.
We’re producing more machine-generated words than at any point in human history.
- AI can write your email.
- AI can write your résumé.
- AI can write your ad.
- AI can write your blog post.
- AI can write your book.
And now some of the companies building those machines are buying old-fashioned physical books, cutting the bindings off, scanning the pages and feeding the words back into AI.
Apparently, after inventing artificial intelligence, Silicon Valley discovered the hottest new technology:
A book written before artificial intelligence showed up.
Welcome to 2026.
THEY’RE REALLY CUTTING UP THE BOOKS
This isn’t some meme somebody invented on X.
In August, 404 Media tracked a shipment containing rare books across the country and reported that it ended up at an Amazon facility in Las Vegas. Workers there described receiving large shipments of printed books, removing the bindings so the pages could be scanned rapidly, and destroying the physical copies in the process.
Why?
AI training data.
And Amazon isn’t the first company whose book appetite has attracted attention.
Court records revealed that Anthropic, maker of Claude, bought millions of physical books, cut them apart, digitized them and discarded the paper originals as part of its effort to build a massive library for AI development.
Anthropic even hired Tom Turvey, who had previously led partnerships for Google’s book-scanning project. According to reporting based on the court record, the mission was essentially to acquire an enormous universe of books for digitization. – Ars Technica
We spent years hearing that AI was going to make books obsolete.
Now AI is shopping for books by the truckload.
You couldn’t write this stuff.
Well, technically ChatGPT could.
But apparently it would like a few million books first.
WHY OLD HUMAN WRITING IS BECOMING SO VALUABLE
Here’s where this gets really interesting.
The internet used to be an enormous source of human-created writing.
Now the internet is also overflowing with writing created, rewritten, summarized, expanded, SEO-optimized, translated, regenerated and recycled by AI systems themselves.
So if you’re trying to train another AI model, there’s something unusually attractive about an old physical book:
You know a human wrote it.
That helps explain the appetite for older material and the sudden interest in giant physical collections.
We built machines capable of producing an almost infinite supply of words.
And then discovered that one of the valuable raw materials for making those machines better is a finite supply of words written by people.
That’s almost beautiful.
In a completely ridiculous dystopian sort of way.
WAIT. THEY CAN LEGALLY DO THAT?
Here’s where it gets even weirder.
In Bartz v. Anthropic, a federal judge ruled in 2025 that Anthropic’s use of copyrighted books to train its AI could qualify as fair use.
The physical-book operation mattered because Anthropic had purchased those books, converted each one into a digital format and destroyed the corresponding paper copy rather than creating an additional copy for distribution. – Reuters
But that wasn’t the end of the case.
Anthropic had also accumulated millions of unauthorized digital books from pirate libraries. The court treated that issue differently, and litigation over those pirated works eventually produced a $1.5 billion copyright settlement, which received final approval in July 2026. – Reuters
So the headline isn’t simply:
“Judge says AI companies can steal books.”
That’s not what happened.
The much stranger reality is that buying a physical book, destructively digitizing that particular copy and using the resulting text internally received different legal treatment from simply downloading millions of unauthorized copies.
Meaning somewhere along the way we arrived at a legal incentive that can make this sentence rational:
Buy the book. Cut the spine off. Scan it. Destroy the paper.
Congratulations.
You have entered the future.
BUT IS DESTROYING A BOOK ACTUALLY… PRESERVING IT?
This is where the X responses got fascinating.
Some people looked at the story and basically said:
What’s the problem?
- If there are thousands of copies of an ordinary old book sitting around and one gets sacrificed so its information can exist digitally, maybe that’s not destruction.
- Maybe that’s preservation.
- And there’s a legitimate argument there.
- Libraries discard books.
- Publishers pulp books.
- People throw books away.
- Warehouses empty inventory.
A digital copy may survive much longer than the particular paperback that supplied it.
But then somebody asks the obvious question:
- What happens when it isn’t just another paperback?
- What if it’s an uncommon edition?
- What if somebody wrote notes in the margins?
- What if there are only a handful of surviving copies?
- What if the physical object itself contains historical information the OCR file doesn’t capture?
And here’s an even better question raised by the discussion:
If a company destroys a physical book to create a valuable digital asset, who gets access to the digital version?
- Everybody?
- Libraries?
- Researchers?
- The public?
- Or just the corporation that bought the book?
Suddenly “digital preservation” becomes a little more complicated.
THERE IS ANOTHER WAY
The funny thing is that none of this proves books have to be destroyed for AI.
Oxford’s Bodleian Library is working with OpenAI to digitize historical, out-of-copyright material while keeping the physical collections intact. The resulting scans are also intended to become publicly accessible. – The Guardian
The Internet Archive has likewise spent years using nondestructive scanning systems.
So there are really two separate questions:
Should historical writing be digitized?
And:
Do we have to destroy the object to do it?
Those are not the same question.
Even Elon Musk apparently noticed.
In July, Musk said he had instructed the SpaceXAI team to preserve rare books and scan them “the hard way” rather than cutting the spines off. – 24/7 Wall St.
That’s right.
We’ve reached the point where Elon Musk has to tell an artificial-intelligence operation:
Guys. Maybe don’t destroy the rare books.
Progress.
😂
DON’T WORRY. ELON HAS A PLAN FOR US TOO.
This gets even funnier when you zoom out.
Musk has also argued that AI and robotics could eventually eliminate enough employment that society should respond with what he calls “Universal HIGH INCOME” — government checks funded in a world where machines produce enormous quantities of goods and services. – Elon Musk Archive
So apparently the future looks something like this:
- The robots do the work.
- The AI writes the books.
- The government sends everybody a check.
And Elon?
Tesla shareholders approved a performance-based compensation package that could ultimately be worth up to roughly $1 trillion if Musk hits a long list of extremely ambitious targets over the coming decade. It isn’t a trillion-dollar paycheck sitting in his checking account; the maximum award depends on Tesla reaching enormous valuation and operating milestones. – Investing.com
But still.
Come on.
😂
Universal high income for you.
Potential trillion-dollar performance package for him.
And please don’t forget to save a few books for the robots.
The future is going great.
AND THEN THERE ARE THE HUMAN WRITERS
Here’s the part that interests me most.
While AI companies are spending enormous resources acquiring human knowledge, human beings are creating books at an astonishing rate.
Bowker data reported by Publishers Weekly shows that more than 3.5 million self-published print and e-book titles carrying ISBNs were released in the United States in 2025.
That’s up from about 2.5 million the year before.
More than 3.5 million. – PublishersWeekly.com, WritersFreak.com
Think about the contradiction.
AI companies are looking backward through libraries trying to find high-quality human writing.
Meanwhile, millions of humans are sitting right here producing brand-new books.
And many of those writers have a completely different problem:
Nobody knows their book exists.
That’s the part of this story I can’t get past.
- The machines are desperate for human writing.
- The humans are desperate for human readers.
And somewhere between those two things sits an algorithm deciding whether anybody sees either one.
I TYPED ONE WORD INTO X
Here’s how I found this whole story.
I was sitting on my phone.
- Watching Colorado get beat by Texas Tech.
- Watching San Diego play Milwaukee.
- Talking to ChatGPT.
- Reading X during commercials.
And I typed one word into X:
BOOKS
Guess what I got?
- AI companies destroying books.
- Arguments about rare books.
- Arguments about artificial intelligence.
- Elon Musk.
- Political posts.
- Financial posts.
- Random old posts.
- People arguing about digital history.
- A discussion about universal income.
And apparently half the known universe.
😂
That’s what an algorithmic feed does.
It decides what BOOKS means for you.
NOW TYPE “BOOKS” INTO WRITERS FREAK
And guess what you get?
Parking Spots about books.
That’s it.
- Not whatever subject an engagement algorithm decided would keep you angry for another four minutes.
- Not somebody else’s interpretation of what you meant.
- Not a casino wheel of trending nonsense.
You search the Writers Freak content.
If the Parking Spot contains what you’re searching for, you can find it.
And that brings us to the funniest part of this entire story.
THIS PARKING SPOT DIDN’T EXIST A FEW MINUTES AGO
- I saw somebody talking about AI companies buying books.
- I started reading the responses.
- I searched X.
- I found more of the story.
- I started talking to ChatGPT.
And…
- We checked the underlying reporting.
- We found the court decision.
- We found Amazon.
- We found Anthropic.
- We found the arguments for preservation.
- We found the arguments against destruction.
- We found Elon.
- We found 3.5 million self-published books.
And while I was watching two games on television…
This became a Parking Spot.
- Now it has a headline.
- It has a story.
- It has sources.
- It can have images.
- It can have video.
- It has its own URL.
And tomorrow, when that X post has disappeared under 500 million new posts about something else, this page can still be here.
- I can update it.
- I can send somebody to it.
- I can link another Parking Spot from it.
And I can go right back to X and join the conversation with something more useful than another 280-character reply.
Five minutes earlier it was somebody else’s post flying through my feed.
Now it’s something I control.
Welcome to Writers Freak.