Skip to main content
Advertisement
Live broadcast
Main slide
Beginning of the article
Озвучить текст
Select important
On
Off

For the first time, artificial intelligence forced Russian publishers to reconsider the rules for selecting manuscripts. After a major setback with the book "The Mythology of Lovecraft" by 23-year-old author Artem Shirkunov, in which signs of neural network generation were found, the MYTH publishing house decided to introduce text verification for the use of AI. It will run in parallel with a similar plagiarism test. Other major market players, Eksmo and Alpina, are already discussing the possibility of such measures. Together with the experts, Izvestia dealt with the new realities of book publishing in Russia.

Human verification

Previously, editors were exclusively involved in proofreading all texts, but now neural networks themselves will be connected to catch traces of neural networks. This was reported to Izvestia in the MYTH.

— Yes, changes are being made to the editorial and publishing process. We are standardizing the verification for the use of AI, which will be conducted in parallel with the verification for plagiarism," the press service of the publishing house replied.

ии
Photo: IZVESTIA/Yulia Mayorova

In addition, the editorial staff increased the time for internal proofreading of manuscripts, updated the regulations for literary and scientific editors, and introduced an additional level of control. If, after the comments of the scientific editor, the author adds new footnotes, citations or other materials, the manuscript will be sent for additional scientific examination.

At the same time, the publisher admits that it is too early to fully trust algorithms. Services sometimes label even the works of classics as generated. Therefore, professional editorial expertise remains the main tool: checking the sources, logic, facts, structure of the text and the reputation of the author.

— We are testing AI detectors, but we keep in mind that so far they are producing unstable and sometimes even absurd results. So, when checking Prishvin's books, the AI detector showed an almost 90% probability of using neural networks. It is also believed that artificially generated text may be excessively metaphorical, but non-fiction authors do not necessarily have to write dryly and without metaphors. Therefore, we are looking for a compromise between technology and common sense and are attracting specialists whom we can trust," the press service of the MYTH said.

книги
Photo: IZVESTIA/Andrey Erstrem

Similar discussions are taking place in other publishing houses. Eksmo admits that manuscripts with signs of using AI have become more common, but so far this is a small part of the general flow. The publisher does not conduct a special check, since existing detectors themselves often make mistakes.

— Our authors, if they use AI, then as an auxiliary tool — for preparation, ideas, structure. It is impossible to calculate the share: there is no reliable verification, and we do not want to organize a witch hunt. However, we are already discussing how to approach this issue carefully, more from the point of view of transparency than control. Since 2025, we have been operating an automatic content analysis system, but its purpose is different — to identify possible violations of Russian law in texts, including books in foreign languages. This helps to reduce legal risks and ensure security for the publisher, authors and readers," Evgeny Solovyov, editor—in-chief of the Eksmo art editorial office, told Izvestia.

ии
Photo: IZVESTIA/Sergey Lantyukhov

"Alpina.Prose has no plans to check the manuscripts for AI or plagiarism, relying on the clause on the personal responsibility of the author in the contract.

— We are not defending theses, and we rely on the professional ethics of the writer. "Alpina.Prose" publishes rather complex prose, the authors write texts for a long time, and interaction with AI is never reduced to a text written by a machine. It can be a search for information, fact—checking, structural and compositional integrity — that is, the kind of work that cannot be caught with any kind of verification," Alpina editor-in-chief told Izvestia.Prose" by Tatiana Solovyova.

книги
Photo: IZVESTIA/Andrey Erstrem

Pavel Podkosov, editor-in-chief of Alpina Non-Fiction, says that the company is still discussing future regulations, but considers it acceptable to use AI as a technical assistant. If the book is entirely written by a neural network, the publisher will not publish such a text.

AI goes into circulation

The reason for the revision of editorial procedures was the hype surrounding the book "The Mythology of Lovecraft", published this spring by the publishing house MYTH. The author was a previously unknown 23-year-old Artem Shirkunov. Shortly after the publication, fans and researchers of Howard Lovecraft's work discovered factual errors in the text, references to non-existent studies, and fictional quotes.

After an internal review, the IIF acknowledged that the book did indeed contain errors and traces of AI generation in the footnotes, list of sources, and indexes. The publisher apologized to the readers and promised to prepare the correct materials.

книга
Photo: Global Look Press/Siegfried Kramer

The Russian case is far from unique. In the spring of 2026, the large American publisher Hachette Book Group refused to release Mia Ballard's horror novel Shy Girl in the United States and withdrew the book from sales in the UK after accusations of large-scale use of AI. According to an investigation by The New York Times, more than 75% of the text was generated by a neural network. Ballard herself denied using AI personally and said the responsibility lay with the hired editor.

Another high-profile episode is related to the victory of 61-year-old Caribbean writer Jamir Nazir in the prestigious Commonwealth Short Story Prize literary competition. The author beat 7,800 participants from 54 countries, but some users ran Nazir's story through an AI detector that installed 100% of neural network generation. After a large-scale check of the drafts, notes, and metadata of the author, it turned out that the text was completely written by a human. Due to health problems, Nazir used an automatic dictation program, which is why neural networks considered the text to be machine-generated.

ии
Photo: IZVESTIA/Sergey Lantyukhov

Until a few years ago, publishers were debating whether artificial intelligence could write a novel. Today, technically, any student can write it with the help of one literate script. But until the last writer's voice is silenced in the world, none of the editors will drive nails into the coffin of publishing.

Переведено сервисом «Яндекс Переводчик»

Live broadcast