The period from 2023 to the days we are living in will probably be remembered by historians of the future as the one in which humanity left the “Gutenberg Galaxy” forever. The Gutenberg Galaxy it is, as is known, a book published in 1962 by the Canadian sociologist Marshall McLuhan in which we reflected on the social implications of the invention of the printing press in 1455. What is happening before our eyes today, however, is humanity’s entry into another environment, which we could call the “Claude Galaxy”, a place full of shining suns, but also of secretly threatening black holes. Claude obviously stands for the homonymous Artificial Intelligence license plate Anthropicbut we could also have said ChatGpt (OpenAi), Grock (Elon Musk), Gemini (Google) and so on.
Let’s start with a fact: according to the data analyzed by Ark Invest2025 saw the historic overtaking of humanity’s great language learning models in terms of the volume of texts produced and published. In other words, robots wrote more than us last year. Making precise estimates is however difficult, also because only a small part of the books generated with artificial intelligence are presented as such. However, it is estimated that the books “written” by the algorithms currently in circulation are between 2 and 5 million (but these are numbers that risk being burned in a few weeks).
Goodbye to Gutenberg: in 2025 robots wrote more texts than humanity
In some sectors, such as “guides to success”, the share of books written with bots would reach 77%. In others, such as health guides linked to the world of herbalism, the artificial share of these texts reaches as much as 82%. The situation is particularly critical on the ebook market. The most alarmist observers speak of over 70% of new self-publishing Kindle suspected of being partially or totally AI-generated.
In addition to the “generalist” chatbots that we have already mentioned, there are also platforms designed exclusively for literary creation. This is the case of PlottyBotan Italian software designed to guide the user in the creation of original books ready for printing or digital distribution. «The user», he explained to askanews Simone Sassolico-founder and chief sales officer of the project, «only needs to insert an idea, such as the plot of a novel or the practical topics of a guide. Think about everything else PlottyBot: develops the idea, writes the text, corrects and perfects until a complete, finished and laid out book is created.” If the volume is successful and in this way you manage to become an established (so to speak…) author, you can always publish your autobiography with Autobiographer, another app designed to transform personal stories about users’ experiences into linked chapters.
The Claude Galaxy and the publishing market flooded with millions of AI books
In this constantly fermenting chaos, there is obviously no shortage of crafty people who ride the wave. A year ago the publishing house Tlon PwC published the essay Hypnocracy – Trump, Musk and the new architecture of reality, by the mysterious Chinese philosopher Jianwei Xun, which was soon translated into various other languages. It was later discovered that in reality the author was the Italian philosopher and publisher Andrea Colamedici. Or rather, he was the inspiration behind it, because the book had been written with artificial intelligence. Discovery, Colamedici he threw it at the “brilliant” provocation, even if it is difficult to consider it little more than a joke. In the US, Hachette recently withdrew one of its novels titled Shy Girl from the market after discovering that software had been used to write it. The author, Mia Ballard, attributed the responsibility to one of her collaborators who was responsible for reviewing the first version of the book.
In an essay published a few weeks ago (Plato in Silicon Valley, Ponte alle Grazie), the philosopher Simone Regazzoni raised the alarm: «When we say that generative AI is a technological tool like others, although more powerful, we are simply exorcising the disturbing revolution that it represents due to its ability to produce symbolic contents indistinguishable from living humans». From the chatbots would emerge a “thought without body and without unconscious, without doubts, without holes, without hesitations, without eros, capable of producing fast and efficient performances”, but also capable of “erasing the singularity of living thought which embodies the responsibility of thought itself – which I assume here on myself, with my signature, my story, my desire, my body, my sex, my life, my encounters, my loves, my unconscious, my death”.
Writers against algorithms: will the “Human Authored” stamp stop the bot invasion?
On a more pragmatic level, the publishing world is looking for possible remedies. THE’Authors Guild The American Writers’ Association launched the “Human Authored” certification program exclusively for its members in January 2025, with the aim of distinguishing books written by human beings from the mass of titles generated by an algorithmic agent. Since March 2026, the program has been open to all US authors, and in parallel the UK’s Society of Authors has started a partnership on the topic. Since its launch, over 3 thousand authors have certified 5 thousand titles. There are also applications like CopyleaksGptZero or QuillBot, capable of examining a text and determining whether it was composed by artificial intelligence. Professors know something about them, and they use them to track down students looking for shortcuts for their homework. But, as with anti-doping which is destined to remain eternally behind doping, the evolution of chatbots is always faster than that of sentinel programs. There are also more empirical methods to identify a text written with AI: when there is an excess of schematization, the repeated repetition of the same concepts, an overflow of long dashes, then warning lights come on. But obviously the identification “by eye” of these stylistic features is even more uncertain.
The bald paradox in publishing: how much AI is needed to erase the author?
Then there is the copyright issue. In the main legal systems of the world, copyright is recognized only for works created by human beings. Furthermore, generative AIs are trained on large quantities of texts taken from the Web in a way that is not always transparent and respectful of intellectual property, so the doubt arises that the user himself, by benefiting from this “training”, is complicit in large-scale editorial piracy. The EU Law on Generative Artificial Intelligence stipulates that its providers must ensure that content produced exclusively by bots is identifiable and that it is clearly and visibly labelled. It remains to be seen whether an “accelerationist” sector like this can maintain its driving force and its undoubted potential even when harnessed to a traditional regulatory framework. Or if, vice versa, there is no risk of being left at a standstill even on more strategic issues due to an excess of prudence.
We must then take into account all the nuances. In addition to texts entirely produced by robots, in which the human simply gives a few initial instructions (the famous prompts), we must in fact consider the increasingly frequent partial use of digital tools: a survey by the BookBub portal on 1,200 authors found that approximately 45% of writers currently use generative models to support their work. Indeed, a platform like Claude it can act as a 2.0 search engine, it can translate texts, aggregate data, create interactive tables, correct papers, make summaries, organize schedules, suggest changes. An invaluable help for anyone who works in the field of cultural production. But as in the baldness paradox of the megarian philosophers (if I remove one hair you don’t become bald, if I remove two the same thing, but then at what number of hairs removed do you become bald?), one wonders: at what level of help provided by AI does the text we are writing begin to become a human-robot co-production? There are no easy answers, but the suspicion arises that those that are too apocalyptic and those that are too enthusiastic are both wrong.
