Translating Books in a Click – AI Dystopian Reality
Amazon did it again. Last month, in November 2025, the multinational launched a new service, Kindle Translate, that can supposedly translate between English and Spanish and from German to English. Currently in beta, this AI-powered translation service allows Kindle Direct Publishing (KDP) authors to translate their ebooks for free. According to Amazon, this service enables authors to reach new audiences and earn more, stating that authors can have their translations ready to sell in a few days. Once the “translation” is ready, another AI tool automatically evaluates it as a “quality check” before publication, allowing authors to preview or directly publish the finished work. These translated books will carry a label stating they are a product of Kindle Translate.
A Well-needed Pushback
Of course, with such a development, there are pushbacks. One of the strongest criticisms comes from professional translators, creatives, and cultural institutions. For example, the European Writers’ Council and the European Council of Literary Translators’ Associations issued a joint statement calling on all actors in the publishing industry (writers, agents, publishers, readers) to stand in solidarity with human literary translators. They urge solidarity to protect what (human) translators do, their craft and skills, and the cultural significance of their role in society. Recognising the limitations of machine translation in capturing nuance, tone, and cultural references helps foster respect for human craftsmanship and cultural sensitivity, both of which are vital to literary integrity.
Money Matters
Let’s start with a fundamental point: AI models have been trained on copyright-protected materials, often illegally: there’s no consent, attribution, or remuneration to the owners of said copyright. It seems paradoxical that some authors are inclined to use this service, even though the mere concept contradicts what they ultimately aim to achieve: earning more money. Their own data trains these models, and the authors receive no compensation from them.
Surprisingly, many claim that this type of service lowers the barrier for indie authors. These authors want to translate their books into several languages to increase their revenue, but the cost of hiring professionals quickly rises and becomes an expense they cannot afford. So, they willingly give away their work for free. In other words, they are shooting themselves in the foot. Others argue that it is part of the process of democratisation of “literature”, but how can surrendering the intellectual property of so many to a handful of for-profit corporations be characterised as democratic?
What might happen
It goes without saying that this represents a significant risk to the future of the writing industry. If authors come to rely on this type of free AI solution, demand for high-quality literary translators will decrease drastically. This can cause many professional translators to considerably lower their fees to “compete” for the little work available. Consequently, there will be a reduction in the overall esteem of the craft, which will, in turn, reduce the number of people entering the profession.
Now, the issue with this might not be so clear, but on the road to democratisation, the opposite might be happening. As the industry moves toward automation, what happens to human work? Will human translation then become a niche privilege available only to high-profile writers or certain prestigious books? Will the translators in charge of those books be the only ones to receive respectable fees? Who even gets the opportunity to pursue translation as a career when most rates can barely make ends meet? Where’s the democratisation there? Eroding the foundation of this profession leaves only elitism in its wake.
Human Touch
Undoubtedly, this decision will deeply affect the publishing industry as a whole, particularly the translation market; many translators and agencies are already losing clients, especially indie authors, due to AI. Yet, these authors are hardly victors themselves. Usually, the translation quality can make or break a book’s success, and by trusting AI to deliver a quality product, authors put their reputations on the line. Readers are not forgiving when they feel cheated or deceived, and why would they be after paying top dollar for an automated and soulless translation? The extensive use of this type of tool diminishes both the demand for professional translators and the overall quality of translated literature.
Limitations of AI
Can AI really capture subtleties? Can AI produce impactful literature? AI-powered translation cannot handle the layers of references and meaning present in literary texts. Without the human touch, the texts lose the nuance that makes literature and translation art forms. Readers deserve to be captivated by stories and language; a “functional translation” is neither enough nor fair; they deserve to experience literary texts in all their glory, with their complexity and style. Emphasising the irreplaceable role of human translators can inspire appreciation for the creative act of translation and its importance in literature. While there can be a place for these tools, applying them blindly to literature is like replacing a hearty artisanal feast with meal replacements.
Faulty AI Model
From a more technical perspective, the large language models (LLMs) commonly known as AI are essentially probability models of languages, where meaning, position, and attention —a mechanism for assigning importance to individual words based on the context of a given task— are encoded and mathematically processed to find patterns over many thousands of sentences. The model learns to predict words based on these encoded variables and consumes accessible data, often from the Internet —and not necessarily legally. These models can then generate content for the Internet, which, eventually, trains new iterations of those very same AI models.
Because of the nature of the algorithms that determine the probability of a given word following a given sentence in a specific context, the content generated will tend toward some “average” response. This ongoing loop means the quality of AI-generated materials tends to degrade over time, as the training data becomes saturated with “average” content, magnifying mistakes and imperfections. If you’re interested, this paper explains this notion in detail. Interestingly, last week, Rockstar cofounder Dan Houser compared this situation to mad cow disease… food for thought.
What about Quality and Accuracy?
While Amazon claims that the translations produced by Kindle Translate are “automatically evaluated for accuracy” before release, there is no information on how this process works or what issues it detects. The author can check the items marked by the system. However, if the author doesn’t understand or lacks the necessary knowledge to evaluate those issues, the product’s quality is clearly in jeopardy, especially in literary texts, whether fiction or nonfiction. Recognising the limitations of automated quality checks can foster trust in professional translators’ expertise and the importance of human oversight for literary quality.
And Transparency? - Ethical Use of AI?
More often than not, the uses of AI are questionable at best (just read this list of AI risks and dangers compiled by IBM, this article on AI risks by the Center for AI Safety, or, if you really want to read all the nitty-gritty, their full paper). Vitally, AI is a convenience tool, much like convenience food; it is far from a replacement for human translation, especially for works of literary merit, whether fiction, nonfiction, poetry, or journalistic pieces.
As AI translation tools become widespread, a rise in works that falsely claim to be human-made might also become increasingly common. Then, how this will affect copyright laws and the integrity of authorship beyond AI is anyone’s guess…
Transparency and responsibility are of utmost importance; authors, agencies, or publishers need to disclose (loud and clear) when translations are AI-generated. By stating how the books come to be, authors enable readers to choose whether to support, read, or consume AI-generated books. This level of transparency is vital to the industry’s survival. Without clear disclosure, readers might unknowingly buy books produced by AI, leading to feelings of betrayal. Discontented readers are not authors’ best friends — we have previously discussed in detail what happened in one of these cases in this article.
Then What?
This AI issue is not just a matter of translation; AI threatens most creative endeavours… the very endeavours that make us human. In the end, literature is more than “convenience food”; it is a form of expression capable of transcending cultures. Ultimately, it is “nourishing food”. While the desire to reach new audiences is laudable, it should not come at the cost of the integrity of the work. If, for cost reasons, the default becomes using AI tools, as a collective, we lose more than we gain: the homogenisation of literature and the underappreciation of the art of translation, instead of the multiplicity of perspectives and the pride of creation. We need to consider where we place our values: Are speed, convenience, and profit more important than the true value of literature and the livelihood of many professionals?
More Questions
One could argue that answering the following questions will become more pressing as AI gains ground:
What is literature? What is translation? Is it just a mechanical act that an AI can do? Or is it a creative act of interpretation, creativity, and re-creation?
At what cost do we prioritise access and reach? There’s nothing wrong with indie authors wanting to reach global audiences, but if the cost of that is providing a low-quality product and bastardising literature, is it worth it? Why do global audiences deserve less than the original readers when they, too, are paying for these works?
What is good enough? And who defines good enough? Which books and sources did the AI train on? How can authors ascertain the quality and legality of those texts and intellectual property?
Only time will tell if the answers we have will hold true in the end.
This post may contain affiliate links
