AI Leaves Translator With No Income: "There's Nothing Now"

From Earning "A Lot" to Zero: AI Leaves a Translator With No Income

More than eight million jobs could be lost in Spain by 2033. This is the figure that the United Nations Conference on Trade and Development (UNCTAD) put in black and white in a 2025 report, and an English and Spanish to French translator is now experiencing it firsthand. Her monthly balance can be summed up in three unadorned sentences: "I used to earn a lot, and now with artificial intelligence, nothing"; "Nothing now, I'm using my savings to live." And a list of expenses that admits no cuts: 1,000 euros for rent —she lives alone— and between 400 and 500 euros for food per month.

The case is not an anecdotal rarity. It is the tip of the iceberg in the translation sector, which has been seeing every software advancement eat away at its fees for years. First, it was computer-assisted translation tools; then, statistical machine translation engines; now, large language models that return plausible text in seconds at bargain prices.

From 250 Euros Per Language to Zero Cost: The Math That No Longer Adds Up

The pogre of prices is the most eloquent aspect. Translating product manuals into several languages used to cost around 250 euros per language. Today, the same work is run through an automatic translator and comes out at zero cost. Those who recount this place it three or four years ago and add that the market has not reversed.

A freelance translator with more than two decades of uninterrupted activity —without a single day of sick leave— describes the same collapse from within. 95% of her clients were agencies from seven or eight countries across four continents. Her specialty, technical translation, used to pay well. For the past one or two years, the client's question has shifted: why spend on quality if the result passes the filter?

The argument that seals the deal is not technological; it's market-driven. If neither the major content platforms nor the big media outlets seem willing to have their texts "reflect" that they spend money on revision, a small or medium-sized enterprise won't be holier than the Pope. Junk food, fast fashion, and acceptance of mediocre quality as the standard: the translator isn't losing to the machine, they're losing to the indifference of the payer.

Pharmaceuticals, Manuals, and Post-Editing: Where AI Already Works Unsupervised

The pharmaceutical sector is cited as the starkest example. Documentation that previously passed through specialized hands is now fed directly into an automatic translator and pushed forward. There are errors, it is admitted, but unless they are very glaring, no one stops to quibble over one word or another. The exception is reports going to the FDA or EMA, where signatures and legal requirements mandate that someone with a name and surname be accountable.

That is the real boundary. As pointed out in the debate, legal responsibility cannot be delegated to a model: any document that must attest, certify, or have legal consequences needs a human signature. It's not a concession to the profession; it's a design limitation. And, for now, it sustains sworn translators.

What Translations Can't a Machine Sign?

Literary translation is another separate territory. Translating a novel is as much an art as it is a transfer of meaning, and the result is noticeable. It's enough to compare a well-translated edition of a classic from the early 2000s with a recent reissue to see how much the standard has degraded in just two decades. The selling price, however, has not decreased.

Something similar happens with simultaneous interpretation. Local colloquialisms, ironies, silences that miccionan something different depending on the interlocutor's culture, body language: a model doesn't process them, and those who have worked with professional interpreters point out that the gap remains enormous. It is the part of the profession where the client pays for judgment, not for words.

The Pattern Already Seen with Photography and Portrait Artists

It's worth looking back. Photography sidelined portrait artists and press illustrators within two or three decades, and that transition affected only one sector. Digital photography wiped out an entire ecosystem: film factories, transportation, developing, retail stores. The difference now is the simultaneity: not one trade falls, but several fall at once.

The Productivity Argument Versus Jobs

The most repeated defense of automation is the usual one, and it's not trivial. If growing potatoes costs 50 euros per ton in labor and a machine brings it down to two cents, the producer loses their business, but everyone else eats cheap potatoes. The uncomfortable question isn't technological; it's about distribution: what happens when the one who loses their salary is also the one who has to buy?

Those who defend this thesis insist that no one proposes competing against a machine, just as no one competes against a car. The car delivers more packages, not fewer. The problem is that here, the car also knows how to read, write, and decide, and the adjustment isn't measured in decades but in months. One forecast summarizes the general sentiment: in two years, not even the recognition of the labor market will be the same.

The translator from the beginning is left with an unanswered question: what does one retrain into when their career was built on two languages and a judgment that the machine, for now, doesn't replicate? No one has yet put a price tag on that.