A few months ago, a bestselling American book arrived in Spanish bookstores after an astonishingly quick translation and revision process. Soon after, controversy erupted among the reading community. TikTok was flooded with videos of readers pointing out odd phrases, typos, and passages that simply didn’t sound like Spanish. The debate was clear: who was behind the translation, an algorithm or a person?
Although the publisher never officially confirmed it, this episode leaves a question hanging in the air: how can a reader, without access to the original or knowledge of translation, know if the book they have in their hands has been translated by a machine?
The short answer is that you can’t always know for sure. The long answer is that there are quite recognizable signs, if you know where to look.
The text is ‘correct’, but it doesn’t sound natural
It would be absurd to deny the immense progress of current machine translation; for example, it no longer makes obvious grammatical errors. The problem lies elsewhere: it generates sentences that, word for word, make sense, but that no native speaker would construct that way. This is the result of translating the original language literally instead of reformulating or substituting with its equivalent in the target language.
A typical example: English tends toward short sentences with minimal punctuation, while Spanish tends toward subordination. When a translation adheres too closely to the short structure of the original, the text sounds choppy, as if it’s running out of breath. Conversely, overly complex sentences with added information not present in the original are signs that the translation is “filling in” rather than reformulating.
Metaphors can be translated, but they are not understood.
Metaphors, idioms, and colloquialisms are closely linked to the culture, humor, and history of a country or community, and they lend fluency and naturalness to a text. They are rhetorical devices with figurative meanings; a word-for-word translation is not feasible because the text would be nonsensical, sound strange, and many nuances specific to the language and culture would be lost.
For example, if English uses an image, such as ” her thoughts were way off track ,” and Spanish translates it literally—”sus pensamientos se salen del carril”—instead of seeking a more natural equivalent—”estaba distraída,” “tenía la cabeza en otro sitio”—the result falls somewhere between understandable and ridiculous. Human translators recognize these expressions and resolve them using translation strategies; AI tends to translate too literally, and as a result, the text loses the naturalness of the original.
The characters change gender without explanation.
English does not mark gender in nouns. When an English text refers to a profession without specifying who holds it—a doctor, a neighbor, a judge—and the translation alternates between masculine and feminine for the same person, it is an almost unmistakable sign of machine translation.
This is not an isolated case: several recent academic studies have confirmed that machine translation systems continue to default to the generic masculine or resolve gender inconsistently within the same text. A human translator resolves this ambiguity by carefully reading the entire work; an automatic system translates phrase by phrase using statistical calculations, and sometimes contradicts itself.
Same object, different terms
When the same word from the original appears translated in different ways—for example, a shop that is sometimes translated as “business,” sometimes as “stall,” and sometimes as “bazaar”—it is usually an indication that no one has reviewed the entire text with a consistent editorial approach. While a professional translator maintains a consistent vocabulary throughout the work, an automated system processes the text in fragments and sometimes forgets its own previous decisions.
Typos that are not simple typos
Generative AI systems sometimes invent words or modify real terms, especially in long or complex passages. This phenomenon is known as ” hallucination “: in order to preserve the coherence of the paragraph, the system invents a word or entire phrase that doesn’t even exist in the original. If a recent edition from a reputable publisher contains typographical errors, unthinkable if it had undergone a professional proofreading process, one must ask which stage of the editorial process failed.
The translator’s name is not shown.
If a book’s credits don’t list the translator’s name, it’s the clearest sign that there’s no human behind the translation. Every professional translation must include the translator’s name, both in accordance with copyright law , which recognizes translators as authors of derivative works, and as a matter of publishing policy. If no name appears, or only the publisher’s name is listed, it’s reason enough to be suspicious.
Why it matters
Individually, none of these signs proves anything definitively. A tired translator or one with a tight deadline might slip in a typo or simply copy a metaphor. But when several of these signs happen to converge in the same book, it leaves no room for doubt.
On social media, the debate has erupted. More and more videos are circulating that openly ask whether AI poses a threat to literary translation. Furthermore, some readers have written their own guides, driven by the unease of not knowing, upon opening a book, who actually wrote it in their language. Both are signs that this concern is no longer confined to the translation profession.
This matters beyond mere reader curiosity; behind each of these details lies a significant editorial decision: how much time and budget are invested in making a book sound good in another language. Literary translation is not a technical process; a book is not just information, it’s rhythm, irony, wordplay, cultural nuances, a nod to the author’s genius and narrative subtlety; and reading it in another language only works if it’s done with care. When that process is accelerated or made cheaper, and we let the machine do everything, the one who pays the price is, quite literally, the one who buys the book.
Literary translation is not simply transferring words from one language to another; it is an act of human expression and creation. Therefore, even though AI is already part of its present, the question is whether publishers will address this with the same clarity with which they protect their copyrights.
Author Bio: Cristina Cela Gutiérrez works Translation and Interpreting at the University of Las Palmas de Gran Canaria
