At some point in the next several months, I am hoping to receive a modest check as a member of the class covered in the class-action settlement Bartz v. Anthropic. In 2025, the artificial intelligence (AI) company Anthropic, best known for creating the chatbot Claude, agreed to pay up to US$1.5 billion to thousands of authors after a judge ruled that the company had infringed upon their copyrights.
When I first learned about the settlement, I assumed that Anthropic was primarily interested in teaching Claude about the subject of my stolen work, former socialist activist, British Labour politician and feminist Ellen Wilkinson.
It did not initially occur to me that Claude might also be learning about how I, Laura Beers, political historian, craft my sentences and translate my voice to the page.
Yet there is increasing evidence that chatbots like Claude can be trained not only to regurgitate an author’s content, but also to mimic their voice. In March 2026, journalist Julia Angwin filed a class action suit against the owners of Grammarly, alleging that the company misappropriated her and other writers’ identities to build its “Expert Review” AI tool, which offers to give editorial feedback in the voices of various authors, living and dead.
That a machine might use my writing not only to learn about my subject matter, but also to analyse and ultimately mimic my authorial voice, points to a future that George Orwell envisioned with eerie prescience. In his 1949 dystopian novel “1984,” Orwell imagined “novel-writing machines” capable of mass-producing literature, employing programmed mechanical “kaleidoscopes” as substitutes for individual artistic process.
‘In place of a human mind, a statistical average’
To what extent, I wondered, could Claude mimic my own voice?
I prompted the chatbot to write an essay “in the style of Laura Beers” about an early 20th-century politician. I struggled to recognise myself in the resulting prose. But maybe Claude, having only consumed one of my books, didn’t have enough to go on to really nail my tone.
I decided to try again. This time, I would ask it to write a brief essay in the style of Orwell. The subject of my book “Orwell’s Ghosts: Wisdom and Warnings for the Twenty-First Century,” Orwell is one of the most distinctive literary essayists and prose stylists of the 20th century.
Most of Orwell’s work is in the public domain, meaning it could have legally been used to train Claude.
I prompted Claude to write an essay “in the style of George Orwell about the threat to individual identity and freedom of thought posed by AI and particularly by the ability of ‘novel writing machines’ to produce literature.”
Primed with ample source material, the AI did a passable job:
“Here is a type of comfort, familiar to anyone who has ever been slowly dispossessed of something, that consists not in denying the loss but in not yet noticing it. The frog does not leap from the gradually heated water. The citizen does not protest the gradual narrowing of the permitted thought. And the reader, it seems, does not mourn the novel until the novel has already been replaced by something that resembles it in every outward particular – same chapters, same characters, same approximate sequence of feeling – yet contains, in place of a human mind, a statistical average of all the human minds that came before it.”
The final sentence about the statistical average rings false. But Orwell would, I suspect, have liked the image of the slowly boiling frog. “Here is a type of comfort” is also a phrase that Orwell might well have written.
I am sceptical that anyone would classify Claude’s efforts as indistinguishable from Orwell’s prose. But when it comes to machine-produced “literature,” perhaps it doesn’t really matter whether it can fully approximate original art, as long as it’s good enough to function as entertainment and distraction for the masses.
Jam, bootlaces and books
This was Orwell’s own dispirited suggestion in “1984.
With the help of “novel-writing machines,” the employees of the Ministry of Truth – the government department responsible for controlling information and rewriting history – are able to mass-produce not only novels, but also “newspapers, films, textbooks, telescreen programmes [and] plays.” They churn out “rubbishy newspapers containing almost nothing except sport, crime and astrology, sensational five-cent novelettes” and “films oozing with sex,” along with cheap pornography intended for the “proles,” as the uneducated working classes of Big Brother’s Oceania were known.
The technology disgusts Orwell’s protagonist, Winston Smith, who pointedly decides to purchase a diary and pen to write down his own independent thoughts. But to Julia, Winston’s nymphomaniac, anti-intellectual lover who works as a mechanic servicing the machines, “Books were just a commodity that had to be produced, like jam or bootlaces.”
‘Full-Length Novels in Seconds’
According to estimates, thousands of books for sale on Amazon have been written in whole or in part using AI.
In other words, today’s AI is also being used to mass-produce literature like jam or bootlaces.
Many of these works are not fully machine-written. Instead, they’ve been, as the AI writing tool Sudowrite advertises, “polished by AI.” With its “Rewrite” function, the company promises to give users an opportunity to “refine your prose while staying true to your style, with multiple AI-suggested revisions to choose from.” The service is akin to the “touching up” provided by the Ministry of Truth’s Rewrite Squad in “1984.”
Other books for sale on Amazon are, however, entirely machine-generated. The AI writing tool Squibler promises that if you give it an overarching prompt, it can produce “Full-Length Novels in Seconds.”
The potential of AI-generated “literature” to turn a quick-and-easy profit ensures that readers will continue to encounter more of this content in the future, especially as AI’s large language models become more refined. Already, studies have shown that readers cannot easily distinguish AI-generated forgeries from original prose.
Last year, I had lunch with a screenwriter friend in Los Angeles. He told me that his colleagues are particularly nervous about the use of AI to produce sequels. Once you have an established cast of characters for a movie franchise like, say, “Fast & Furious,” audiences will likely see the next instalment, whether it’s written by man or machine.
Yet my own brief experiments with Claude give me at least some hope for the future of literary art. A chatbot like Claude might be able to absorb and analyse “a statistical average of all the human minds that came before it,” but without the input of actual human experience and sensibility, it is hard to envisage them ever producing true art.
Whether AI can produce the next George Orwell novel or essay remains to be seen. That it can and will churn out an increasing volume of popular fiction and screenplays like “Fast & Furious 25” seems less in doubt.
- is a Professor of History, American University
- This article first appeared in The Conversation
