This profession deserves better

What “unsustainable” actually means. A profession, by definition, is the activity you earn your living from. The day translation can only sustain those who are just starting out, or topping up another income — a salary elsewhere, a partner's earnings, a pension — there will be no profession left to speak of. Not only because nobody could live on this work any longer, but because nobody could stay in it long enough to build the experience it demands. Mastery in translation is not issued with a diploma: it accrues over years in the field, and it lives in the people who practise it. Take away the possibility of a career, and you take away the seniors. The juniors who still start out will find no senior left to ask, to learn from, to be corrected by — only the old posts those seniors once left in forums and blogs, preserved somewhere on the internet like exhibits in a museum. That is what extinction looks like for a knowledge profession: not an empty room, but a room where the expertise is no longer alive. And it is not a distant scenario: in a January 2025 survey by the UK's Chartered Institute of Linguists, members describe winding down towards early retirement, or watching their remaining work degrade into low-value post-editing.¹

The costs you pay. Running a translation business means an ever-growing list of hard costs: hardware, software licences, an expanding stack of subscriptions, insurance, taxes, professional memberships — and a workspace, whether rented or carved out of your home and your family's patience. Recent research on self-employment found that independent workers spend 5–8 percentage points more of their gross income on unreimbursed expenses than employees do.²

The time you lose. Beside the costs you pay sit the costs no invoice will ever show: your hours. Every project drags a tail of unpaid work — quoting, invoicing, project setup, style guides, query forms, implementing everyone else's queries, final QA, "small favours" for the client, one agency portal after another, and the marketing that never ends. A 2026 survey of freelancers measured an average of 204 hours a year lost to admin and paperwork — five full working weeks, unpaid. And that is the general freelance economy; a translator's overhead list is longer.³

The revenues that no longer cover either. In 2020, CSA Research surveyed over 7,000 linguists worldwide: half of professional translators were earning less than $20,000 gross a year; one in five, less than $5,000.

Since then, every needle has moved the wrong way. In 2024, for the first time in the history of the European Language Industry Survey, most freelance translators reported falling rates — despite inflation. In 2025, only 57% said their freelance work earns them enough to live on. In the same period, 7 out of 10 freelancers surveyed by the Chartered Institute of Linguists saw their work volumes shrink — for half of them, significantly — and 43% of translators surveyed by the Society of Authors reported income lost to generative AI.⁶⁷¹⁵

The equation. Costs rising, sellable hours shrinking, revenues falling: three curves that only need to hold their course for the sum to reach zero.

Where the jobs come from. A translator has four ways to reach the market. A personal website — a shop window on a street nobody walks down: it can amplify an opportunity, never generate one. A profile in a directory, waiting its turn among thousands of near-identical pages. Agencies, where the rate arrives already decided: take it or leave it, negotiable only for rare languages and emergencies. And the platforms where demand actually meets supply — through a mechanism that deserves a closer look.

Reverse auction. The pattern is familiar to everyone. A job is posted, dozens of applications pour in within minutes, and the client ends up choosing by the only criterion they can actually understand. No platform is built or deliberately run to hand jobs to the lowest bidder; none rewards undercutting — some even explicitly warn translators and clients about the risks of prices that are too low. But that wall of offers comes with nothing to tell one from another, so price remains the one difference that catches the eye. And so, without anyone directing it that way, the outcome is still that of a reverse auction — the procurement format created for one declared purpose: driving the price downn (online reverse auctions — a research review, ScienceDirect).

Why the slope tips one way. Market power has always favoured the buyer; an open, unregulated profession, an increasingly interchangeable service (see Problem 3) and the arrival of AI (see Problem 4) keep shifting it further. Researchers now call this the "uberization" of translation: studies document falling rates, diminished bargaining power, growing dependency on platforms — and translators with no voice in how those platforms are run (Fırat 2021, John Benjamins · "Translators in the platform economy", Perspectives 2024).

Over a decade on the record. In 2012, a European Commission study already described a market in disorder: traditional ways of standing out no longer working, online visibility turning into something anyone can buy, and a warning that good translators may leave the market altogether (Pym, Grin, Sfreddo, Chan — The Status of the Translation Profession in the European Union, DGT 2012). Twelve years later, an analysis of five million Upwork listings measured the slide from inside the channel itself: translation postings down 19%, hourly rates down more than 20% — the steepest drop of any category (Bloomberry, "The jobs being replaced by AI — an analysis of 5M freelancing jobs", 2024).

The spiral is not a forecast. It has been turning for over a decade — and it is picking up speed.

Scholars put it in writing fifteen years ago. In 2012, a study commissioned by the European Commission examined how translators signal their worth to the market — and found the signals failing. Its conclusion: when buyers cannot tell a good translator from a bad one, they pay everyone below their real value.

On one side, a demand that cannot judge. This is an industry where the client mostly has no idea of the value of what they are buying — so even those able to spend more can see no reason to. The Commission's study explains the phenomenon:

On the other, a supply — overabundant and undifferentiated. With no barriers to entry and no signal able to separate those who say they are translators from those who are, any of the two billion bilingual people in the world can start selling translations, on their own terms and at their own price, styling themselves a professional. And the professional has no working way to prove otherwise:

No entry barrier, no price floor. When anyone can walk in and undercut, prices keep sinking — and every drop squeezes out precisely those who make their living at professional rates. Some snapshots taken along the way:

The 2012 study used the future tense. Today's surveys use the present.

The signals, one by one. Look at every tool a translator has to prove their worth, and behind each one sits the reason it cannot be trusted. Try it yourself: imagine you have been handed a generous budget to hire the best translators in the world. How would you find them?

The proof? Signals so weak that stealing them has become a trade. Scammers lift the CVs of reputable translators, swap in their own contact details, and mass-mail them to unsuspecting agencies; clients hire, the text goes through machine translation, the payment vanishes — and the blame lands on the real translator, whose CV and reputation were taken. The numbers: one watchdog group reports over 22,000 email addresses tied to translation scams, growing by more than a hundred every month; a single UK company received 2,173 fake CVs in one year, one Belgian translator 893, and each stolen CV is mailed out to two or three thousand recipients. The existence — and the scale — of this trade leads to a bitter conclusion: today's signals have grown so weak they no longer even stick to their owner. They circulate like bearer instruments, in a market that seems to have lost interest not only in value, but in truth itself.³⁴

The ground that is gone. The numbers are not in dispute. Post-edited machine output runs, by industry estimates, around 80% cheaper and up to ten times faster than translating from scratch. And the textbook example comes from the buyer's side, in the buyer's own words: a leading language-learning company needed about twelve years to build its first hundred courses — and, having moved its translation and content work to generative AI (cutting, in two rounds, the contractors who used to do it), launched 148 new courses in under a year.

Twelve years of work, done in one: that is AI's ground. Whoever sells good-enough translation is standing on it — as it shifts.

What the machine cannot sell. Now look at what the buyers do next — because it is the most instructive fact in the whole story: in 2025 industry research, among those using machine translation or large language models, 90 to 98 percent still apply human post-editing to the output before using it.

Three years into the AI era, almost nobody trusts it alone — because fluency hides failure: machine output reads smoothly even when it is wrong. What buyers are still paying humans for is no longer the typing. It is the two things no machine can deliver about its own work: the assurance that the text is actually right, and the excellence that good-enough, by definition, is not — the critical document, the rare language pair, the text where a mistake costs more than the translation.

More machine output, more to verify. Here is the part of the story nobody tells: efficiency does not shrink markets — it floods them. Economists have known it since 1865, when more efficient steam engines increased Britain's coal consumption (the Jevons paradox). It happened to a sister profession in real time: in 2016 a founding father of AI declared it was time to stop training radiologists; imaging volumes have grown ever since. And economists now name translation among the occupations where AI-driven productivity can expand demand for the human function rather than erase it. The mechanism is already visible: AI multiplies what gets translated — and everything it translates is a candidate for verification. The demand for the human layer grows with the machine's output.⁵

A demand that cannot find you. Growing — and bought blind. The profession's mass response to AI has been to stay on the machine's ground as its discounted supervisor: in a 2025 industry survey, 88% of freelancers reported working in machine-translation post-editing, and 86% said its rates have worsened; reported discounts cluster around 20% and reach 29%. The industry, meanwhile, names post-editing its strongest growth target — the same service that posted the worst performance of all in 2024.

This is what demand looks like when it cannot find you: the human layer is needed everywhere, but with no way to tell who actually guarantees a text from who merely clicks through it (see Problem 3), it is bought as a commodity, at auction (see Problem 2). And the stake is arithmetic: the work that remains — assurance and excellence — either reaches the few hundred thousand professionals this market actually counts (estimates run from the ~330,000 of the European Commission's study to the millions of self-declared profiles online), or it dissolves among two billion bilinguals, at two-billion-bilinguals prices.

This profession has buried its own obituary before. Translators have been declared obsolete at every technological turn since the first translation engines went online — and many of those declared obsolete went on to build careers and raise families — and are grandparents now. This wave is bigger. It will be navigated all the same — but not with the old utopias: no register, no licence, no global union has ever fit a borderless market. What can work is the opposite of dispersion: a core of professionals who become individually recognizable and collectively present, instead of atomized — and who stop working against themselves. This page is a first attempt at exactly that: gathering the profession around the profession's own interests. Perhaps what gets built from here will keep that unity of purpose alive.